AM Edition: Top 10 Security Intel Articles on LiveNews.co.nz for April 22, 2026 – Full Text

Source: MIL-OSI-Submissions-English

Source: The Conversation – UK – By Andreas Krieg, Associate Professor, Defence Studies Department, King’s College London

Following the last round of talks between the United States and Iran in Islamabad, Iran’s foreign minister and negotiator Abbas Araghchi declared in a post on X on April 17 that the Strait of Hormuz was “completely open”. This came after he also signalled that his government could be flexible over the issue of nuclear enrichment as well as Iran’s support for its proxies in the region.

Then came an abrupt correction. Mohammad Bagher Zolghadr, a former commander in the Islamic Revolutionary Guards Corps (IRGC) who was recently appointed as secretary of the Supreme National Security Council, is understood to have complained to the IRGC, submitting a report that criticised Araghchi for “deviation from the delegation’s mandate”.

The negotiating team was called back to Tehran. Araghchi was attacked by state-run media which said his post had “provided the best opportunity for Trump to go beyond reality, declare himself the winner of the war and celebrate victory.” And the Strait of Hormuz was declared closed.

This episode demonstrates the new reality in the Islamic Republic, where the IRGC increasingly calls the shots in all matters of statecraft and government. The rest of the state is a façade at most.

Over the six weeks of war, Iran’s former leadership has been decimated: the supreme leader, Ali Khamenei, was killed in a US strike on the first day of US and Israeli attacks. Many of his senior colleagues have also been killed. Iran is no longer best understood as a state with a powerful militia. It has become, more precisely, a powerful militia with a state – a political order with the IRGC at its core.

The other traditional centres of power – the government and the clergy – have effectively been relegated to mere front organisations. Amid the fog of war, even the new supreme leader, Mojtaba Khamenei, appears merely as a legitimising ornament. In any case, Khamenei is reported to have been severely injured in the attack that killed his father and is apparently taking no part in government.

So who is running the country? The answer points unmistakably to the IRGC and its leader, Ahmad Vahidi.

Guardians of the revolution

The IRGC was created after the 1979 revolution, precisely because Ayatollah Ruhollah Khomeini and his allies did not trust the conventional state apparatus to defend the revolution. Over time it grew beyond its role as guardians of the revolution into an all-encompassing, all-channel network. It became a military, an intelligence service, an economic conglomerate and a regional expeditionary network. Its internal security force, the Basij, gave it an arm of mass social control inside Iran. The Quds force was set up to export the revolution across Iran’s proxies in Lebanon, Iraq, Syria, Yemen and beyond.

Far from destroying this architecture, sanctions deepened it. They led to the creation of front companies linked to the IRGC doing illicit deals and operating circuits of patronage that enriched those closest to the centre of power. What emerged was a parallel state that gradually outgrew the formal one.

The IRGC is organised as a network with a core and a periphery. Its central hub decides strategy. This is surrounded by a network of decentralised cells capable of operating with a high degree of autonomy. This is called Iran’s “mosaic defence doctrine”. And it was built to operate precisely the way it is now: to keep fighting amid attempts at decapitation and disruption.

A new leader emerges

After IRGC chief Mohammad Pakpour was killed on the opening day of the conflict, Ahmad Vahidi, a former interior minister and a founding member of the IRGC, has emerged to take his place. After being appointed in an emergency capacity after his predecessor was killed, he has consolidated effective control as the civilian presidency has been hollowed out.

With the new supreme leader apparently incapacitated and the clergy sidelined, Vahidi and his group of allies – IRGC commanders and security council hardliners such as Ali Akbar Ahmadian and Mohammad Bagher Zolghadr – have set the mandate and red lines for the ceasefire talks.

The IRGC’s red lines are clear: it will not surrender uranium enrichment altogether; it wants to preserve its missile program and the axis of resistance; it wants sanctions to lifted and access to Iranian assets overseas that are presently frozen. Room for negotiation only exists on technical details about enrichment levels, timelines for lifting sanctions or the language of any deals that are agreed.

In times of war, states tend to centralise as civilian institutions shrink. Hard men tend to rise, especially after many of the influential political pragmatists, such as Ali Larijani, the former secretary of the security council, were deliberately taken out by Israel.

The IRGC was not suddenly conjured by this war, but prepared by decades of institutional entrenchment, economic capture and delegated coercion. The IRGC’s military dictatorship in the making needed this war to consolidate its influence over competing nodes in the network – most importantly the clergy.

This has profound consequences for the negotiations. Instead of being straightforward bargaining between statesmen, Washington’s real estate moguls turned negotiators are speaking to Iranian counterparts who are on a short lead held by the IRGC. Progress in negotiations should not be judged by what Iran’s diplomats say in public, but by what the guard allows to be implemented in practice.

Trump and Israel’s failed decapitation strategy leaves a potent system in place that feels emboldened by the desperation in the White House to find a diplomatic off-ramp. To think that this war-hardened system of hardliners will capitulate is wishful thinking.

The past few days have made it clear that the IRGC is now a militia with a state using the civic and military institutions of the Islamic Republic as its outer skin. While there is room for negotiation to build a mutually acceptable deal, the US administration needs to be realistic about where the IRGC’s red lines are and what card it actually has to play against a resilient network with a very high threshold for pain.

Andreas Krieg does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

– ref. Who is calling the shots in Iran? – https://theconversation.com/who-is-calling-the-shots-in-iran-281066

Source: MIL-OSI-Submissions-English

Source: The Conversation – USA – By Anne Toomey McKenna, Affiliated Faculty Member, Institute for Computational and Data Sciences, Penn State

On a Saturday morning, you head to the hardware store. Your neighbors’ Ring cameras film your walk to the car. Your car’s sensors, cameras and microphones record your speed, how you drive, where you’re going, who’s with you, what you say, and biological metrics such as facial expression, weight and heart rate. Your car may also collect text messages and contacts from your connected smartphone.

Meanwhile, your phone continuously senses and records your communications, info about your health, what apps you’re using, and tracks your location via cell towers, GPS satellites and Wi-Fi and Bluetooth.

As you enter the store, its surveillance cameras identify your face and track your movements through the aisles. If you then use Apple or Google Pay to make your purchase, your phone tracks what you bought and how much you paid.

All this data quickly becomes commercially available, bought and sold by data brokers. Aggregated and analyzed by artificial intelligence, the data reveals detailed, sensitive information about you that can be used to predict and manipulate your behavior, including what you buy, feel, think and do.

Companies unilaterally collect data from most of your activities. This “surveillance capitalism” is often unrelated to the services device manufacturers, apps and stores are providing you. For example, Tinder is planning to use AI to scan your entire camera roll. And despite their promises, “opting out” doesn’t actually stop companies’ data collection.

While companies can manipulate you, they cannot put you in jail. But the U.S. government can, and it now purchases massive quantities of your information from commercial data brokers. The government is able to purchase Americans’ sensitive data because the information it buys is not subject to the same restrictions as information it collects directly.

The federal government is also ramping up its abilities to directly collect data through partnerships with private tech companies. These surveillance tech partnerships are becoming entrenched, domestically and abroad, as advances in AI take surveillance to unprecedented levels.

As a privacy, electronic surveillance and tech law attorney, author and legal educator, I have spent years researching, writing and advising about privacy and legal issues related to surveillance and data use. To understand the issues, it is critical to know how these technologies function, who collects what data about you, how that data can be used against you, and why the laws you might think are protecting your data do not apply or are ignored.

Big money for AI-driven tech and more data

Congressional funding is supercharging huge government investments in surveillance tech and data analytics driven by AI, which automates analysis of very large amounts of data. The massive 2025 tax-and-spending law netted the Department of Homeland Security an unprecedented US$165 billion in yearly funding. Immigration and Customs Enforcement, part of DHS, got about $86 billion.

Disclosure of documents allegedly hacked from Homeland Security reveal a massive surveillance web that has all Americans in its scope.

DHS is expanding its AI surveillance capabilities with a surge in contracts to private companies. It is reportedly funding companies that provide more AI-automated surveillance in airports; adapters to convert agents’ phones into biometric scanners; and an AI platform that acquires all 911 call center data to build geospatial heat maps to predict incident trends. Predicting incident trends can be a form of predictive policing, which uses data to anticipate where, when and how crime may occur.

DHS has also spent millions on AI-driven software used to detect sentiment and emotion in users’ online posts. Have you been complaining about Immigration and Customs Enforcement policies online? If so, social media companies including Google, Reddit, Discord, and Facebook and Instagram owner Meta may have sent identifying data, such as your name, email address, phone number and activity, to DHS in response to hundreds of DHS subpoenas served on the companies.

Meanwhile, the Trump administration’s national policy framework for artificial intelligence, released on March 20, 2026, urges Congress to use grants and tax incentives to fund “wider deployment of AI tools across American industry” and to allow industry and academia to use federal datasets to train AI.

Using federal datasets this way raises privacy law concerns because they contain a lifetime of sensitive details about you, including biographical, employment and tax information.

Blurring lines and little oversight

In foreign intelligence work, the funding, development and controlled use of certain AI-driven gathering of data makes sense. The CIA’s new acquisition framework to turbocharge collaboration with the private sector may be legal with proper oversight. But the line between collaborating for lawful national security purposes versus unlawful domestic spying is becoming dangerously blurred or ignored.

For example, the Pentagon has declared a contractor, Anthropic, a national security risk because Anthropic insisted that its powerful agentic AI model, Claude, not be used for mass domestic surveillance of Americans or fully autonomous weapons.

On March 18, 2026, FBI Director Kash Patel confirmed to Congress that the FBI is buying Americans’ data from data brokers, including location histories, to track American citizens.

As the federal government accelerates the use of and investment in AI-driven spy tech, it is mandating less oversight around AI technology. In addition to the national AI policy framework, which discourages state regulation of AI, the president has issued executive orders to accelerate federal government adoption of AI systems, remove state law AI regulation barriers and require that the federal government not procure the use of AI models that attempt to adjust for bias. But using advanced AI systems is risky, given reports of AI agents going rogue, exposing sensitive data and becoming a threat, even during routine tasks.

Your data

The surveillance capitalism system requires people to unwittingly participate in a manipulative cycle of group- and self-surveillance. Neighborhood doorbell cameras, Flock license plate readers and hyperlocal social media sites like Nextdoor create a crowdsourced record of all people’s movements in public spaces.

Sensors in phones and wearable devices, such as earbuds and rings, collect ever more sensitive details. These include health data, including your heart rate and heart rate variability, blood oxygen, sweat and stress levels, behavioral patterns, neurological changes and even brain waves. Smartphones can be used to diagnose, assess and treat Parkinson’s disease. Earbuds could be used to monitor brain health.

This data is not protected under HIPAA, which prohibits health care providers and those working with them from disclosing your health information without your permission, because the law does not consider tech companies to be health care providers nor these wearables to be medical devices.

Legal protections

People have little choice when buying devices, using apps or opening accounts but to agree to lengthy terms that include consent for companies to collect and sell their personal data. This “consent” allows their data to end up in the largely unregulated commercial data market.

The government claims it can lawfully purchase this data from data brokers. But in buying your data in bulk on the commercial market, the government is circumventing the Constitution, Supreme Court decisions and federal laws designed to protect your privacy from unwarranted government overreach.

The Fourth Amendment prohibits unreasonable search and seizure by the government. Supreme Court cases require police to get a warrant to search a phone or use cellular or GPS location information to track someone. The Electronic Communications Privacy Act’s Wiretap Act prohibits unauthorized interception of wire, oral and electronic communications.

Despite some efforts, Congress has failed to enact legislation to protect data privacy, the use of sensitive data by AI systems or to restore the intent of the Electronic Communications Privacy Act. Courts have allowed the broad electronic privacy protections in the federal Wiretap Act to be eviscerated by companies claiming consent.

In my opinion, the way to begin to address these problems is to restore the Wiretap Act and related laws to their intended purposes of protecting Americans’ privacy in communications, and for Congress to follow through on its promises and efforts by passing legislation that secures Americans’ data privacy and protects them from AI harms.

This article is part of a series on data privacy that explores who collects your data, what and how they collect, who sells and buys your data, what they all do with it, and what you can do about it.

Anne Toomey McKenna serves on the Advisory Board to the Institute for Electrical and Electronics Engineers (IEEE)-USA’s Artificial Intelligence Policy Committee (AIPC) and Chairs multiple AIPC subcommittees. The AIPC work involves subject matter and education-related interaction with U.S. Senate and House congressional staffers and the Congressional AI Caucus. McKenna has received funding from the National Security Agency for the development of legal educational materials about cyberlaw (a course which the government still makes available online for the public) and funding from The National Police Foundation together with the U.S. Department of Justice-COPS division for legal analysis regarding the use of drones in domestic policing.

– ref. US government ramps up mass surveillance with help of AI tech, data brokers – and your apps and devices – https://theconversation.com/us-government-ramps-up-mass-surveillance-with-help-of-ai-tech-data-brokers-and-your-apps-and-devices-277440

Source: MIL-OSI-Submissions-English

Source: The Conversation – UK – By Anja Shortland, Professor in Political Economy, King’s College London

When evolutionary biologist Joseph Popp coded the first documented piece of ransomware in 1989, he had little idea it would become a major criminal business model capable of bringing economies to their knees.

Popp, who worked for the World Health Organization at the time, wanted to warn people about the dangers of ignoring health warnings, poor sexual hygiene and (human) virus transmission.

He sent out 20,0000 floppy discs that, when loaded, flashed up a demand for money to regain files that had supposedly been encrypted (in fact, it was just their file names). He was later arrested and charged with 11 counts of blackmail, but declared mentally unfit to stand trial.

In 1996, two Columbia University computer scientists published a paper explaining how criminals could use more sophisticated versions of Popp’s scheme to mount large-scale extortion operations. At the heart of this was malicious software that could be used to encrypt, block access to or steal a person or organisation’s files and data.

However, two preconditions still had to be met for ransomware to become a feasible criminal business: communication channels that were difficult to monitor, and a payments process outside financial regulation.

The Tor protocol, released by US intelligence services to protect their covert communications, solved the first problem in 2004. Cryptocurrencies solved the second – in particular, when bitcoin cash machines started appearing in North American cities from 2013.

Today, artifical intelligence makes malware coding and crafting convincing phishing-emails in any language simple. And the latest model in Anthropic’s AI system, Claude Mythos, recently proved more effective at hacking into computer systems than humans.

As an expert in extortive crime, I am increasingly concerned about public and political apathy to the threats posed by ransomware. To better understand these, it’s worth tracing its evolution over the past two decades – and how improvements in computer security and law enforcement, plus changes in data regulation, have led to new criminal strategies each time.

Cut out the middlemen

The first generation, which came to global attention in the mid-2010s, was known as “commodity ransomware”. A pioneering example, Cryptolocker, was developed by Russia-based hackers who infiltrated hundreds of thousands of computers, seeking to cut out the middlemen previously needed to commit financial fraud. They proved that a large majority of their victims would happily pay a small ransom to restore data that had been locked by their malware.

As both competent and incompetent hackers piled into this new market, victims shared information about rogue operators and put them out of business. This led to the second generation of ransomware such as Ryuk, which emerged in 2018.

In this phase, criminals abandoned the indiscriminate “spray-and-pray” approach in favour of targeting individual cash-rich businesses. They would set an individual ransom, negotiate with the company, and even offer to help with decryption if paid. Fast-rising ransoms more than compensated for this increased administrative effort.

In response, many companies began investing in multi-factor authentication, better threat monitoring, advance warning systems and software patches for known vulnerabilities.

However, these security benefits were soon offset by the impact of COVID on work practices across the world. The pandemic led to widespread remote working, with many people using unsecured devices and connections that were vulnerable to cyber-attack.

A multibillion-dollar industry

The next ransomware innovation was driven by the emergence of back-up systems that enabled companies to restore encrypted files without the criminals’ help. This was coupled with the emergence of tighter data privacy regulation such as GDPR in Europe and the UK.

Invented in 2019, third-generation ransomware weaponised these regulations, which threatened firms with massive fines if confidential data about clients or staff was revealed. The criminal gangs now sought out and exfiltrated an organisation’s most sensitive files, then threatened to publicise them through dedicated dark web leak sites.

This so-called double-extortion model – encrypting an organisation’s data while threatening to make it public – brought many businesses back to the negotiation table.

Ransomware had become a multibillion-dollar industry – with the Conti gang, sheltered by Russia and employing hundreds of people, among the key players setting new records for ransomware demands. Its attacks on critical infrastructure and hospitals saw it sanctioned by the UK government in 2023.

This new approach forced many governments to row back on imposing hefty fines for data breaches, since many were the result of criminal attacks. Meanwhile, new initiatives by law enforcement – supported by the private sector – targeted and broke up the largest and most egregious ransomware gangs.

Today’s fourth generation of ransomware, building on the latest AI technology, looks nimbler and slimmed-down in comparison. Anyone who gains access to a network can lease weapons-grade malware on the dark web without forming long-term ties with a particular gang.

Advanced AI-based hacking tools make ransomware accessible to many more criminals and politically motivated hacktivists. And around one-quarter of breaches still result in ransom payments. For criminals sheltered by their governments, only the digital infrastructure is at risk of being taken down by western law enforcement.

Lessons not learned

While coverage of Claude Mythos suggests even the most sophisticated cyber defences could now be vulnerable, the troubling reality is that many individuals and organisations are still using out-of date, unpatched or only partially upgraded software. This means even early-generation ransomware techniques are still lucrative.

While Popp sent out his floppy discs to promote better sexual hygiene, today’s poor cyberhygiene is leaving many public and private networks open to malware attacks. The intended lesson of his original ransomware caper – be vigilant and properly heed health warnings – has still only been partially learnt in the digital world.

Many western societies appear to have grown accepting of criminals leaching on business conducted on the internet. Not even a steady stream of human fatalities, caused by attacks on hospitals and medical providers, has generated the level of response required to stamp out this dangerous threat.

The hope that governments sheltering cybercriminals can be encouraged (or forced) to stop them targeting critical national infrastructure appears increasingly fragile amid current geopolitical tensions. At all levels of society, we need to get smarter about cyber defence.

Anja Shortland does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment. Anja’s latest book, We Know You Can Pay a Million: Inside the Dark Economy of Hacking and Ransomware, is published by Profile Books.

– ref. From floppy discs to Claude Mythos, how ransomware grew into a multibillion-dollar industry – https://theconversation.com/from-floppy-discs-to-claude-mythos-how-ransomware-grew-into-a-multibillion-dollar-industry-281000

Source: MIL-OSI-Submissions-English

Source: The Conversation – UK – By Andrew Gawthorpe, Lecturer in History and International Studies, Leiden University

When Donald Trump deported a group of Venezuelan nationals to El Salvador in 2025, it was the fulfilment of a long-held wish. Across both of his administrations Trump has pushed officials to find ways to brutalise immigrants, particularly those who are undocumented, believing that doing so will deter others from making the trip.

The Venezuelan nationals were destined for El Salvador’s Terrorism Confinement Center, known as Cecot. When they arrived, according to a Human Rights Watch report, they were subjected to systematic beatings, sexual abuse and psychological duress.

The Trump administration amplified reports of conditions in the prison. Trump’s former homeland security secretary, Kristi Noem, for example, filmed a video inside Cecot in 2025 in which she thanked El Salvador for “bringing our terrorists here and incarcerating them”.

Trump’s deportations were a chilling sign of how easy it is for US presidents to sidestep the constitution. If Cecot were in the US, it would be recognised as a site of illegal abuses. The constitution’s protection against “cruel and unusual punishments” would cause judges to order it shut down – and it is likely that political outrage would not cease until that order was followed.

Yet by making an agreement with El Salvador’s president, Nayib Bukele, Trump managed to get around these legal and political obstacles. In a recent paper, I explored how Trump’s deportations are part of a broader pattern of what I call “presidential extra-territorialization” – American presidents acting in or through a foreign jurisdiction to circumvent the US constitution.

There is a long-term pattern of cooperation between presidents from both the Republican and Democratic parties and the leaders of foreign countries. It is a pattern that could have grave implications for the future of US democracy.

Outsourcing abuses

The ability of US presidents to engage in this outsourcing of abuses is rooted in two things. First, their control over the vast capabilities of the modern executive branch, with its array of spies, soldiers and law enforcement officials. And second, control over US diplomacy, which is enshrined in Supreme Court precedent.

In 1936, the court ruled that the president is “the sole organ of the federal government in the field of international relations”. This has commonly been interpreted as meaning US presidents cannot be constrained by the other branches of government when conducting diplomacy.

Combined, these factors mean presidents face fewer constraints in foreign affairs than in the domestic realm. They are able to avoid oversight from the courts and Congress by keeping agreements with other governments secret and by acting too fast to be stopped. If they can find just one foreign government willing to enable them, then what is not possible at home suddenly becomes possible overseas.

This lack of constraint was evident in Trump’s deportations. The US government sent the men to El Salvador despite a last-minute ruling by a federal court ordering their return.

And once they were in El Salvador, the Trump administation claimed it was no longer responsible for them and could not be expected to bring them back. The Supreme Court stepped in to pause further such deportations, but only weeks after the fact.

Other examples of the power and flexibility of extra-territorialization became apparent during the “war on terror”, when successive US presidents faced the issue of where to send detainees who were suspected terrorists.

If they were brought to the US, they would have had constitutional rights and could not have been tortured or indefinitely imprisoned. So presidents from Bill Clinton in the 1990s onward established a series of agreements with other countries to take and mistreat them instead.

After the 9/11 terrorist attacks in 2001, the Bush administration established a series of “black sites” in countries such as Poland, Thailand and Romania in which to hold detainees in secret. Abuses were committed directly by US agents, but still beyond the reach of US courts. The administration held prisoners at Guantanamo Bay in Cuba too, another place where the constitution’s reach was limited.

Presidents can also shift territory in response to attempts to constrain their actions. When the US Supreme Court ruled that detainees at Guantanamo Bay had to be afforded certain rights in 2008, the Obama administration transferred some detainees to Bagram Air Base in Afghanistan. Bagram was not covered by the Supreme Court ruling.

As a US court of appeals noted in 2010, the ability to shift territories so easily seemed to allow the administration to “switch the constitution on or off at will”.

Yet another example of extra-territorialization is the “Five Eyes” intelligence agreement between Australia, Canada, New Zealand, the UK and US. As part of this pact, members have been reported to spy on each other’s citizens – an outsourcing of surveillance that allows each to circumvent domestic privacy constraints.

The fact that Trump has engaged in extra-territorialization so openly, in contrast to previous administrations who tried to keep it hidden, is a stark warning.

Even when the president said he was exploring a proposal to send US citizens to Cecot in April 2025, he received little pushback from within his own party. This suggests they have accepted it as a legitimate strategy to achieve policy goals.

In the hyper-polarised atmosphere of contemporary US politics, extra-territorialization is threatening to become a regular tool of governance. To stop that from happening, it is vital to expose and confront it. But first we must understand it.

Andrew Gawthorpe is affiliated with the Foreign Policy Centre in London.

– ref. How US presidents shift controversial actions abroad to get around limits at home – https://theconversation.com/how-us-presidents-shift-controversial-actions-abroad-to-get-around-limits-at-home-280769

Source: MIL-OSI-Submissions-English

Source: The Conversation – France – By Frédéric Fréry, Professeur de stratégie, CentraleSupélec, ESCP Business School

OpenAI, the creator of ChatGPT, is gearing up to launch its Initial Public Offerings (IPO) this year. This financial manoeuvre would represent a pivotal shift for a project originally designed for the “common good” towards a market-driven logic. Established in 2015, OpenAI started out amidst growing anxiety regarding artificial intelligence (AI). Founded by Sam Altman and Elon Musk, the tech company adopted a non-profit structure and made no secret of its goal to develop AI that is “beneficial to humanity” and prevent it from remaining in the hands of a few dominant players.

This ambition distinguished it from tech giants like Google, Microsoft, Meta, and Amazon, which were built on proprietary models and rent-seeking effects.

In contrast, OpenAI intended to champion general public interest by emphasising open research and sharing knowledge. However, this orientation – symbolised by its name – quickly collided with a structural constraint: the astronomical cost of generative AI.

Massive costs

Unlike traditional software, where marginal costs tend towards zero (for example, the millionth copy of Windows costs Microsoft nothing), generative AI requires massive infrastructure.

Every interaction mobilises computing resources, energy, and specialised equipment. A standard ChatGPT query, consisting of one question and one answer, costs between $0.01 and $0.10. Similarly, generating a high-definition image can cost between $0.10 and $0.20. While these amounts seem negligible in isolation, they become staggering when scaled to the billions of daily queries seen in 2026.

This is explained by the underlying infrastructure, particularly the Graphics Processing Units (GPUs) supplied by players like Nvidia. These chips can cost tens of thousands of dollars to purchase and several dollars per hour via cloud access.

OpenAI, like its competitors, depends on tens of thousands of these GPUs running continuously in massive data centers. According to some estimates,the necessary investments will reach hundreds of billions by the end of this decade.

As early as the late 2010s, it became clear that a purely non-profit model could not meet such capital intensity. This is why OpenAI adopted a hybrid status in 2019, allowing it to raise funds while maintaining control through a foundation. It was a first foray into the market economy, albeit one tempered by the ambition to resist investor demands.

Brutal acceleration with ChatGPT

However, at the end of 2022, the chatbot ChatGPT radically changed the game, attracting 100 million users in just two months, before surpassing 900 million weekly users by early 2026.

OpenAI’s revenue surged from approximately $200 million (€173.15 million) in 2022 to over $10 billion (€8.65 billion) in 2025 – a sixty-fold increase in three years.

This exponential growth was accompanied by the implementation of a business model with multiple revenue streams. For individuals, OpenAI offers paid subscriptions (ranging from $20 to $200 per month). However, the bulk of the revenue comes from enterprises, via subscriptions priced between $25 and $60 per user per month. A company with 10,000 employees thus represents several million dollars in annual revenue.

Corporate money

OpenAI additionally bills for the use of its models by companies that integrate them directly into their own solutions. Every use is metered, often on a massive scale. An application processing a million queries a day can generate tens of thousands of dollars in monthly billing.

Finally, a growing portion of revenue comes from strategic agreements, notably with Microsoft, which integrates OpenAI technologies into its products under the Copilot brand.

It is the sum of these flows – subscriptions, licences, third-party usage, and partnerships – that allowed OpenAI to reach approximately $1 billion in monthly revenue in 2025. Yet, this commercial rise masks an intrinsic economic fragility.

A gigantic cash-burning machine

Despite sharply rising revenues, OpenAI remains structurally loss-making. In the first half of 2025, the company reportedly generated approximately $4.3 billion in revenue while recording losses between $7 billion and $13 billion – more than $2 billion in losses every month. In total, cumulative losses could exceed $140 billion (€121.19 billion) between 2024 and 2029.

This drift is explained by the very nature of OpenAI’s business model, where every interaction incurs a cost alongside gargantuan necessary investments. Beyond infrastructure, Research and Development (R&D) is a major expense. To stay in the technological race against an increasingly competitive environment, OpenAI reportedly invested nearly $16 billion in R&D in 2025 alone.

To this is added the cost of human resources, which is sometimes extraordinary. While base salaries for the most in-demand AI experts range from $250,000 – $700,000 per year, their total compensation – including stock and bonuses – frequently exceeds $1 million. In some cases, annual compensation even exceeds $10 million. Here again, bidding wars from competitors like Meta force OpenAI to match these offers for fear of seeing its key talent vanish.

Nearing bankruptcy?

In short, OpenAI’s business is not enough to cover its costs, to the point that some analysts suggest that at this rate, it could be forced to file for bankruptcy as early as 2027. Recourse to external financing is therefore indispensable to cover these losses.

To sustain its growth, OpenAI has already raised approximately $58 billion since its inception, including more than $13 billion from Microsoft. In 2025, an exceptional funding round reportedly raised up to $40 billion more, pushing its valuation to several hundred billion dollars.

At the end of March 2026, a new $122 billion funding round – notably involving Amazon ($50 billion), Nvidia, and SoftBank ($30 billion each) – brought the valuation to $852 billion (€737.6 billion). Yet, these amounts remain insufficient given the requirements.

Industrial Dependency

Dependency on industrial partners appears particularly problematic. Microsoft provides OpenAI with its cloud infrastructure via Azure, while Nvidia plays a key role upstream by providing GPUs. Much like the Gold Rush era, when shovel sellers grew rich at the expense of prospectors, it is the infrastructure providers in the AI sector making a fortune, not the model designers.

In practice, every AI query generates revenue for infrastructure providers, amounting to a form of “invisible tax” captured upstream.

In 2025, Nvidia generated nearly $73 billion in net profit on approximately $130 billion in revenue, and its stock market valuation is 1.5 times higher than the entire Paris stock exchange!

Governance missteps

OpenAI’s economic tensions have spilled over into its corporate governance. The hybridisation of a public interest mission with private financing mechanisms resulted in a complex structure. A non-profit foundation controls a for-profit “public benefit corporation”, which is funded by investors and tasked with raising capital and developing activities – all while theoretically remaining subordinate to the foundation’s public interest mission. This construction, designed to avoid purely financial logic, quickly fuelled tensions between different stakeholders.

Elon Musk’s departure in 2018 was the first signal of a strategic disagreement. In 2020, several researchers left OpenAI to found Anthropic, citing differences over safety and governance. However, it was primarily the crisis of November 2023 that fully revealed the system’s fragilities, when the board of directors suddenly announced the firing of Sam Altman, citing a lack of transparency in his communications.

Within hours, the situation spiralled into an open crisis. Nearly all employees threatened to leave the company if Altman was not reinstated. Microsoft, the main partner and investor, publicly supported Altman and even discussed the possibility of hiring him and his teams. Faced with this pressure, the board was forced to reverse its decision within days. Sam Altman was reinstated, and the board’s composition was profoundly overhauled. This episode highlighted internal tensions, specifically the difficulty of making divergent logics coexist within the same company: ethical posturing, industrial imperatives, and investor demands.

Intensifying Competition

In addition to these internal constraints, competitive intensity is particularly fierce.

Google, the inventor of generative AI, is making rapid progress with Gemini. Anthropic, with Claude, has established itself in certain segments, particularly programming, while emphasising safety.

China’s DeepSeek has claimed to use less expensive processors. France’s Mistral AI advocates for a frugal approach and European digital sovereignty. In a sign of this shifting landscape, Apple which initially partnered with OpenAI to include ChatGPT for certain Siri features – has chosen to replace it with Gemini.

In this context of ecosystem reorganisation, OpenAI’s position, while still central, is being challenged. Intensifying competition reinforces the need for ever-greater financial resources.

The stock market: lifeline or mirage?

OpenAI’s Initial Public Offering (IPO) is presented as a response to these constraints: a way to fund massive investments and consolidate a weakened competitive position. An IPO could raise between $50 billion and $100 billion by selling 10% to 20% of the capital. Such an operation would constitute one of the largest in the history of financial markets.

However, this transformation involves delicate trade-offs. A listed company is subject to profitability and transparency requirements that may clash with the experimental nature of artificial intelligence. Added to this is the persistent dependence on Microsoft and Nvidia, which limits the company’s strategic autonomy.

Most importantly, there is no indication that an IPO would suffice to resolve OpenAI’s structural problems. At best, without a significant shift in the business model, it would only delay its bankruptcy by a few years. The economic model of generative AI remains fundamentally unstable today.

A Question Beyond OpenAI

Beyond the case of OpenAI, one can legitimately question the current functioning of an economy dominated by tech giants. Artificial intelligence is establishing itself as an essential infrastructure whose effects far exceed the economic sphere. For some analysts, control over AI now carries the same geostrategic importance link please as the possession of nuclear weapons.

Consequently, a civilisational question arises: can we entrust the development and direction of such a technology solely to financial markets? Can we imagine Elon Musk or Mark Zuckerberg personally owning the equivalent of one or more atomic bombs? OpenAI’s IPO will not provide the answer alone. However, it will constitute one of the first large-scale tests.

---

A weekly e-mail in English featuring expertise from scholars and researchers. It provides an introduction to the diversity of research coming out of the continent and considers some of the key issues facing European countries. Get the newsletter!

---

Frédéric Fréry ne travaille pas, ne conseille pas, ne possède pas de parts, ne reçoit pas de fonds d’une organisation qui pourrait tirer profit de cet article, et n’a déclaré aucune autre affiliation que son organisme de recherche.

– ref. OpenAI gets set to go public: can we entrust the financial markets with ChatGPT and AI? – https://theconversation.com/openai-gets-set-to-go-public-can-we-entrust-the-financial-markets-with-chatgpt-and-ai-280943

Source: MIL-OSI-Submissions-French

Source: The Conversation – France in French (3) – By Louis Perez, Chercheur post-doctorant en droit international, Université Paris-Panthéon-Assas

Le conflit en cours en Iran montre la dépendance avancée des armées états-unienne et israélienne à l’IA militaire, notamment pour le ciblage et la planification des frappes. Le bombardement d’une école à Minab, le 28 févier, présenté comme une erreur de ciblage et ayant causé la mort de 168 civils, principalement des enfants, met en lumière les risques juridiques, les failles des systèmes et les problèmes de responsabilité.

---

Le conflit armé contre l’Iran lancé le 28 février dernier par Washington et Tel-Aviv a été rapidement qualifié de « première guerre de l’IA ». Une assertion en réalité trompeuse à divers égards. Non seulement l’IA a-t-elle déjà été utilisée de façon intensive dans des conflits récents, notamment par Israël à Gaza, mais, plus largement, l’IA, en tant que moyen numérique de traitement et d’analyse de données, entretient une longue histoire avec les conflits armés, dont les fondements techniques remontent à la Seconde Guerre mondiale.

Certes, la situation iranienne se distingue par le niveau de sophistication sans précédent de ces moyens et par la dépendance inédite des armées à leur égard. Elle diffère également du conflit à Gaza en ce que, cette fois, l’IA a été déployée contre un adversaire étatique dans le cadre d’une guerre de haute intensité. Enfin, jamais les États n’avaient aussi ouvertement communiqué sur leur recours à ces systèmes. C’est cette communication conjuguée aux conséquences dramatiques de certaines frappes qui invite à s’interroger sur la compatibilité de ces pratiques avec le droit international.

Les faits : l’utilisation de l’IA dans la guerre en Iran

L’utilisation de l’IA par Israël dans sa guerre contre le Hamas avait été révélée par le journal +972. Ce média avait exposé ce que bon nombre de spécialistes soupçonnaient depuis quelques années. Dans le cadre du conflit en Iran, en revanche, ce sont les autorités américaines elles-mêmes qui ont annoncé leur emploi de l’IA.

Effectivement, les forces militaires américaines ont admis avoir recouru à des systèmes d’IA pour établir et trier la liste des objectifs à une vitesse fulgurante. Ce procédé aurait entraîné plus de 1 000 frappes, qualifiées de très précises, durant les vingt-quatre premières heures du conflit. Elles se seraient notamment servies du système Maven Smart System, un projet conjoint utilisant un logiciel d’IA de surveillance et de collecte de données de Palantir, couplé au système d’IA générative Claude, développé par Anthropic.

Cependant, lors du premier jour de guerre, l’une des frappes américaines a visé une école de Minab, causant la mort d’environ 170 victimes civiles, principalement des enfants. Les États-Unis ont reconnu leur responsabilité dans cette frappe, présentée comme une erreur. L’école était en effet située à proximité d’une base navale des Gardiens de la révolution. Elle faisait autrefois partie intégrante du même complexe, avant d’en être séparée. C’est donc une information non actualisée qui aurait conduit à autoriser la frappe.

Une telle méprise n’est pas anodine. De nombreux médias et ONG ont rapidement établi le lien entre l’école et la base navale. Il a ainsi été avancé que l’armée américaine aurait probablement ciblé ce bâtiment sur la base de données obsolètes en suivant aveuglément une recommandation issue d’un système d’IA sans procéder à la vérification qui s’imposait.

La licéité de l’utilisation de l’IA

Dans quelle mesure l’utilisation de l’IA pour mener ces frappes, et l’erreur commise, sont-elles licites au regard du droit international ?

Il convient d’abord de préciser que l’IA n’est pas interdite en tant que telle par le droit des conflits armés (DCA, également appelé droit international humanitaire). Pour l’heure, aucune règle juridique n’envisage spécifiquement la question de sa licéité. Pour autant, la question n’évolue pas dans un vide juridique. Les règles générales du DCA s’appliquent à la conduite des hostilités, quels que soient les moyens et méthodes déployés.

L’une de ces règles est le principe de distinction selon lequel seules les cibles militaires peuvent faire l’objet d’attaques, les personnes civiles et les biens civils devant être préservés. Viser directement une école, comme celle de Minab, en l’absence de tout objectif militaire en son sein, constitue donc une violation manifeste de ce principe. Il est toutefois peu probable que l’armée américaine ait eu l’intention délibérée de détruire l’école en tant que telle. Comme indiqué, il s’agit plus vraisemblablement d’une erreur d’identification de la cible, possiblement liée à un système d’IA entraîné sur des données obsolètes, datant de l’époque où le bâtiment était encore rattaché à la base navale.

En conséquence, la violation est plutôt relative au principe de précaution. Ce dernier prescrit notamment que les parties au conflit doivent faire tout ce qui est pratiquement possible pour vérifier que les objectifs à attaquer sont bien des objectifs militaires. En l’espèce, l’armée américaine ne semble pas avoir procédé aux vérifications nécessaires pour s’assurer que la cible était une école. Une vérification élémentaire, comme celle effectuée par certains médias, aurait pu rapidement dissiper le moindre doute.

Il faut rappeler que, lors de la guerre à Gaza, il avait été rapporté que des soldats israéliens ne disposaient parfois que de vingt secondes pour valider une cible, ce qui interroge sur la possibilité matérielle de respecter effectivement ce principe. Les préoccupations relatives à l’IA militaire se cristallisent souvent autour de la question de l’autonomie et du risque qu’un système désigne et engage seul une cible ; c’est l’enjeu des systèmes d’armes létales autonomes. Cet exemple démontre toutefois qu’un contrôle humain formellement maintenu peut n’être que fictif si l’opérateur ne dispose ni du temps ni de l’esprit critique nécessaires pour évaluer une recommandation algorithmique.

Du côté iranien, il y a lieu de relever que le principe de précaution n’a pas davantage été respecté. Ce principe impose non seulement des obligations à l’attaquant, mais requiert également de l’attaqué qu’il prenne certaines précautions passives : les parties doivent notamment éloigner les personnes civiles et les biens de caractère civil des objectifs militaires. En l’espèce, transformer un bâtiment d’une base navale en école, tout en la maintenant à proximité immédiate du reste du complexe militaire, exposait délibérément cette installation civile aux risques liés à la conduite des hostilités.

Quelles responsabilités juridiques, politiques et morales ?

La responsabilité individuelle. L’attaque ne constitue pas un crime de guerre.

Si l’attaque constitue une violation du DCA, il est probable qu’aucun militaire américain ne soit condamné pour de tels faits. Outre les questions de compétence juridictionnelle, le principal obstacle tient à ce que ni la violation du principe de précaution ni les erreurs conduisant à des violations du DCA ne constituent des crimes de guerre au sens du droit international pénal.

L’acte matériel est bien caractérisé, mais l’élément intentionnel, c’est-à-dire la volonté de commettre l’infraction, fait défaut. Le régime de responsabilité pénale internationale actuel ne reconnaît pas la responsabilité pour négligence dans ce contexte. Cette approche pragmatique pourrait néanmoins évoluer. D’une part, si les erreurs algorithmiques de ciblage se multiplient, le caractère « raisonnable » de l’erreur sera de plus en plus difficile à invoquer et l’utilisation consciente d’un système connu pour ses défaillances pourrait induire une forme d’intention indirecte de cibler des civils. D’autre part, le droit pourrait à l’avenir se développer pour sanctionner les militaires qui, par leur négligence, causent la mort de civils.

La responsabilité des entreprises d’IA. Un bras de fer entre puissances économique et politique.

Un autre point de vigilance tient au rôle des entreprises privées spécialisées en IA qui détiennent aujourd’hui la majeure partie des compétences technologiques mobilisées sur le champ de bataille. Ces entreprises pourraient être tenues responsables lorsqu’elles développent des systèmes défaillants, mais, au-delà de cette responsabilité, une question morale et politique fondamentale se pose au regard de la vente de technologies d’IA à des fins militaires.

Juste avant l’entrée en guerre des États-Unis, Anthropic, qui produit le système Claude, s’était opposé à une coopération sans limite avec le Pentagone, notamment sur les armes autonomes, en invoquant ses engagements éthiques et les limites de fiabilité technique de ses systèmes pour les usages envisagés. Le Pentagone avait alors accusé Anthropic de trahison, bien que ses systèmes continuent d’être utilisés par l’armée.

D’autres entreprises du secteur, comme OpenAI, Google, Amazon ou Microsoft, semblent, quant à elles, collaborer sans réserve avec les armées, s’imposant de facto comme de véritables entreprises de défense. Il est intéressant de noter que des entreprises, normalement guidées par le profit, ont parfois plus d’états d’âme en la matière que certains États pourtant garants de l’intérêt général.

La responsabilité étatique. Répondre de ses actes et prévenir les prochaines violations.

Les États qui développent et utilisent l’IA militaire portent une responsabilité particulière. En l’espèce, les États-Unis engagent leur responsabilité internationale pour la commission d’un fait internationalement illicite. Cette responsabilité sera, certes, difficile à mettre en œuvre dans la pratique. Mais au-delà, il émerge une responsabilité tant juridique que politique. Aux termes de l’article 1 commun aux conventions de Genève, les États ont en effet l’obligation de respecter et de faire respecter le DCA. Or, le développement de l’IA militaire tend à miner ce respect, voire à favoriser et à dissimuler les violations du droit.

Divers mécanismes pourraient endiguer ce phénomène, comme la formation des militaires aux spécificités des systèmes d’IA, l’élaboration de règles d’engagement propres à l’IA, des garanties techniques de fiabilité et de transparence des systèmes ainsi que des tests et évaluations réguliers. Plusieurs initiatives internationales appellent à intégrer de telles mesures au sein de nouveaux instruments juridiques. Pourtant, la volonté politique fait défaut, notamment chez les États à l’avant-garde du développement et de l’utilisation de l’IA militaire.

Ainsi, Pete Hegseth, secrétaire à la défense des États-Unis, semble en réalité agir dans le sens contraire. Il a récemment limogé des conseillers juridiques militaires qu’il considérait comme des entraves à la bonne conduite des hostilités et a qualifié les règles d’engagement de stupides. Plus largement, les États-Unis s’opposent à toute réglementation juridique internationale de l’IA militaire. L’IA apparaît ainsi comme à la fois l’un des moteurs et le révélateur d’une érosion profonde du DCA.

Jacques Lacan disait : « Le réel, c’est quand on se cogne. » L’accident de Minab constitue un évènement dramatique qui confirme les risques sur lesquels les experts en IA militaire alertent depuis plusieurs années et qui aurait dû susciter bien davantage de réactions.

En réalité, cette information semble avoir été éclipsée par d’autres considérations perçues comme plus urgentes et plus visibles dans le cadre de cette guerre, à commencer par le risque nucléaire. L’accident de Minab n’aura pas été l’électrochoc attendu pour inciter les États à s’entendre sur un cadre juridique spécifique applicable à l’IA militaire. Il reste à savoir si un tel électrochoc est encore possible ou même souhaitable.

Louis Perez ne travaille pas, ne conseille pas, ne possède pas de parts, ne reçoit pas de fonds d’une organisation qui pourrait tirer profit de cet article, et n’a déclaré aucune autre affiliation que son organisme de recherche.

– ref. L’utilisation de l’intelligence artificielle dans la guerre en Iran : que dit le droit international ? – https://theconversation.com/lutilisation-de-lintelligence-artificielle-dans-la-guerre-en-iran-que-dit-le-droit-international-280562

Source: MIL-OSI-Submissions-English

Source: The Conversation – Africa – By Clement Sefa-Nyarko, Lecturer in Security, Development and Leadership in Africa, King’s College London

Ghana’s parliament ratified the country’s first lithium mining agreement in March 2026. This came three years after lithium mining was confirmed as commercially viable in September 2023.

The Ewoyaa Lithium Project, in the Central Region of Ghana, covers an area where farming communities have lived for generations. It spans several communities.

The agreement is between the government and Barari DV Ghana Limited, the local subsidiary of Australia-based Atlantic Lithium. Lithium is a mineral used in batteries that power electric vehicles, renewable energy storage systems and everyday electronics. It’s at the heart of global minerals supply chains to decarbonise energy and transport.

With the deal in place, formal discussions will begin with mining communities about relocation, compensation and restoring livelihoods. Compensation could include payment for land, crops, construction work and other assets that will be affected by mining operations, as required under Ghana’s Minerals and Mining Act.

The ratification of the deal also marks the end of a legal moratorium set out in Ghanaian law. This comes into force once minerals of commercial value are discovered.

The moratorium, which lasted three years in the case of the Ewoyaa Lithium Project, was designed to protect both the state and mining firms from complications such as speculative construction, sudden land claims, and inflated compensation demands that may arise from new developments.

Under Ghana’s mining law, once minerals of commercial value are confirmed, temporary restrictions are placed on new permanent structures, farm expansion and other major land use changes in the affected area. It lasts until there is a mineral agreement and compensation arrangements are clear. The intention is to stabilise land use and ensure fair valuation.

It has profound social consequences.

For people already living in these areas, the moratorium can mean extended periods of uncertainty. During this time, everyday decisions about livelihoods, housing and the future are placed on hold.

Its practical impact is that residents living on or near the mining area can’t build, expand their farms, or make other major decisions about land use.

The affected communities live in a state of suspended time during the moratorium. Farmers are unable to plan their next season confidently. Families delay home improvements. Young people postpone major life decisions because their future access to land remains unclear.

The mining agreement doesn’t end the waiting. Instead, it opens a new phase of negotiations, compensation assessments and administrative back and forth. It could stretch on for months or even years.

This prolonged uncertainty causes real social and economic harm. Yet its effects are often overlooked.

My academic work examines governance, natural resources, politics, and energy transitions. In a recent paper, based on extensive fieldwork in the lithium-rich communities of Ewoyaa, Krampa Krom and Krofu, I investigated how these delays and uncertainty shaped everyday life. I gathered firsthand accounts of how people navigated this period of waiting. All are affected by the project.

The effects were unmistakable. People described the moratorium as a form of “frozen time”, when life could not move forward.

The economic setbacks and emotional strain from long periods of uncertainty often go unrecognised in public policy discussions.

Time on hold

My research identified a number of negative effects of the delays in getting mining operations off the ground.

Firstly, households described how it eroded local opportunities and contributed to young people leaving the area. Young people expressed frustration as their job prospects remained frozen, and they lacked clarity on whether future employment at the mine would be accessible or meaningful.

Many young adults, already frustrated by years of stalled prospects, had left in search of work elsewhere.

The few lower-paid jobs associated with early stage mining activities were not yet available.

Secondly, farmers reported clear losses: they could not expand or invest.

Thirdly, women traders, many of whom sell farm produce and foodstuffs, reported disruptions in household income patterns because farming activities were stalled.

Fourth, community elders, reflecting on years of limited communication, described a growing distrust towards government institutions and the processes governing the mineral agreement.

Across these accounts, what united residents was the feeling that their lives had been interrupted by forces far beyond their control. The moratorium did more than pause development, it suspended decision making, aspirations and the ability to plan even the simplest aspects of the future.

“Time on hold” shaped economic choices, social relationships and the very rhythm of community life.

In my study, I argue that these prolonged delays are a form of “temporal injustice”. This concept emerged directly from listening to residents describe how their aspirations, livelihoods and sense of security were reshaped by bureaucratic time.

Temporal injustice occurs when certain groups bear unfair burdens of waiting, uncertainty and delayed decision-making. These disruptions may seem minor when viewed from the outside. But they have broader implications. They affect project timelines, investor confidence, and the long-term reliability of the supply chains that power the global clean energy transition.

Looking forward

As Ghana and the mining company move into the compensation and community engagement phase, they have an opportunity to address not only material losses but the temporal burdens that communities have endured.

First, compensation frameworks should recognise that the moratorium itself caused harm. Beyond land, crops and structures, policymakers must account for the economic and social costs of years spent waiting.

Second, community engagement must be timely, transparent and genuinely participatory.

Information should flow consistently, especially when people’s livelihoods depend on it.

Third, Ghana should incorporate temporal justice principles into mining governance, including clearer timelines, regular updates and support for communities facing prolonged delays.

Finally, as Ghana deepens its role in the global critical minerals supply chains, local communities should share the benefits rather than being left to carry hidden costs. A just energy transition demands fair distribution not only of mineral wealth, but of time, certainty and opportunity.

Clement Sefa-Nyarko receives funding from UK Research and Innovation (UKRI) for a Future Leaders Fellowship that is researching justice in critical minerals governance and energy transitions. Clement also does occasional consultancy for Participatory Development Associates for research and evaluation in Africa, but not directly related to mining.

– ref. Ghana’s mining law aims to stop speculation but leaves communities in limbo – insights from a lithium case study – https://theconversation.com/ghanas-mining-law-aims-to-stop-speculation-but-leaves-communities-in-limbo-insights-from-a-lithium-case-study-279594

Source: MIL-OSI-Submissions-English

Source: The Conversation – UK – By Ursula F Ott, Professor of International Business, Nottingham Trent University

It is now almost a decade since the UK voted for Brexit and since the tariffs of US president Donald Trump’s first term increased global trade frictions. Brexit removed the UK from the European single market for goods and services. Now though, the country is proposing a pivot back towards alignment with EU regulations.

What could have not been widely predicted back in 2016 was the COVID pandemic, nor a war on European soil. The UK has been exposed to these shocks without the EU support system. So what may once have been impossible to imagine is now on the cards: adopting EU single market rules under new UK legislation.

In May 2025, the UK and EU reached a new trade agreement, paving the way for both sides to move closer on their economies and business. This was hastened by unpredictable US trade tariffs and a weakening of the US-UK-EU relationship. In addition, it has been estimated in a comprehensive study that Brexit has reduced the size of the UK economy by 6-8%.

Politically, the approach announced by the UK prime minister, Keir Starmer, is a courageous step. UK legislation would allow the country to adopt new EU laws without the need for parliament to vote each time. But any plan is certain to provoke strong opposition from the Conservatives and Reform UK.

However, it is a signal of the seriousness of the UK’s intentions to move closer to the EU by adapting to its regulations and giving up independence from EU law. That is a costly move for the UK in terms of its credibility, but the U-turn should reinforce its commitment to the EU.

But beyond this, there are three clear benefits to the UK.

1. The EU is built on rules and regulations that guide the bloc’s labour market, trade and security systems. Alignment would clearly help UK businesses, consumers and individual workers to manoeuvre within these systems.

2. By breaking from the single market, the UK chose a costlier approach to trading and investing across the EU border. Aligning regulations would reduce cross-border bureaucracy.

3. The EU is looking for new trading partners after supply chain disruptions from COVID and the Ukraine war – not to mention the current impact on oil and gas supplies. The EU does not need to rely on the UK, but a new direction in the relationship could reduce the threat of supply chain disruption in future.

A better deal for consumers?

So what could this mean for UK businesses and consumers? Food producers trading within the UK-EU zone would have a quicker turnaround of their fresh produce. This would reach shop shelves in the UK and EU more quickly, giving shoppers better-quality fresh foods.

Reducing the amount of complex paperwork and export health certificates at borders would allow a free flow of fresh food even between Great Britain and Northern Ireland (which remained part of the single market). This trade has been disrupted since Brexit and affects both trade between food producers due to paperwork and border delays, and food security.

Border checks, paperwork and adapting to legal requirements are expensive and so increase food prices (and with that, inflation). Bringing trade between the EU and the UK closer could reduce these costs, and should also allow producers to benefit more from global value chains.

US tariffs are at their highest levels since the second world war, and the knock-on cost effects of supply chain disruption in the Middle East make a strong case for strengthening ties between neighbours.

Going forward, it will be resilience rather than efficiency in trade that will be important for both businesses and nations. Both will want to be able to reconfigure networks at speed. If inflation rises due to product shortages, governments have limited fiscal space to offer direct support to citizens (which would mean increased levels of spending), or to cut taxes.

Another benefit could come in the form of foreign direct investment into the UK from overseas. In 2025, this began shifting from low-cost developing countries towards capital-intensive and technologically-driven investments in developed countries – and especially in the EU (Germany, Italy and France).

Alignment with EU regulation could give investors more confidence to commit to the UK. Foreign direct investment in renewable energy and AI products, for example, would benefit both the UK’s workers and its consumers.

This is a time of new geopolitical alliances, cooperation and blocs. Trading and investment options could help secure economic, political and societal stability in a volatile world. So far, this is a relatively small step by the UK – but starting to align to EU regulations could ease a complex relationship.

Ursula F Ott does not work for, consult, own shares in or receive funding from any company or organisation that would benefit from this article, and has disclosed no relevant affiliations beyond their academic appointment.

– ref. With talk of closer EU alignment, the UK is signalling to Europe that it’s a partner worthy of trust – https://theconversation.com/with-talk-of-closer-eu-alignment-the-uk-is-signalling-to-europe-that-its-a-partner-worthy-of-trust-280961

Source: MIL-OSI-Submissions-French

Source: The Conversation – France (in French) – By Hugo Spring-Ragain, Doctorant en économie / économie mathématique, Centre d’études diplomatiques et stratégiques (CEDS)

L’intelligence artificielle ne détruit pas tant des emplois qu’elle modifie profondément les compétences nécessaires pour les accomplir. De cette confusion entre emploi et compétences risquent de naître des erreurs dans les politiques d’accompagnement des mutations en cours.

---

Chaque grande vague technologique a produit son lot de prédictions contradictoires sur l’emploi. L’intelligence artificielle (IA) ne fait pas exception. Mais avant de savoir combien d’emplois l’IA va créer ou détruire, il faudrait s’accorder sur ce qu’elle automatise réellement. La réponse oblige à distinguer trois notions que le débat public confond régulièrement : l’emploi, la compétence et la tâche.

Les grandes vagues d’automatisation ont suivi une logique remarquablement stable en deux siècles : vapeur, électricité, robotique industrielle ont déplacé les tâches physiques répétitives et épargné le travail cognitif non routinier. Cette régularité empirique a été formalisée par Autor, Levy et Murnane dès 2003 sous le nom d’« hypothèse de polarisation des tâches ».

Une illusion persistante

L’automatisation ronge les emplois intermédiaires, ceux des cols bleus qualifiés et employés de bureau exécutant des tâches routinières, mais épargne les deux extrémités. D’un côté, les tâches manuelles non routinières, comme la plomberie ou les soins, de l’autre, les tâches cognitives non routinières, comme l’analyse, le conseil ou la rédaction experte. Ces dernières constituaient le cœur des professions du tertiaire qualifié, et la conviction s’était solidement installée qu’elles resteraient hors d’atteinte.

---

À lire aussi :
Pourquoi l’IA oblige les entreprises à repenser la valeur du travail

---

Cette conviction reposait sur une confusion conceptuelle qu’il faut dissiper avant tout. Ce n’est pas l’emploi de juriste ou d’analyste financier qui était protégé, c’est un ensemble de tâches précises qui composaient cet emploi et qui résistaient jusqu’ici à l’automatisation. La distinction entre ces trois niveaux est fondamentale.

Un emploi désigne un poste occupé dans une organisation, avec un contrat, un salaire, une fiche de poste. Une compétence est une capacité cognitive ou technique mobilisable dans plusieurs contextes professionnels. Une tâche est une action précise, délimitable, dont on peut évaluer si elle est ou non automatisable à un coût donné. C’est à ce troisième niveau que se joue réellement la transformation en cours, et c’est précisément ce niveau que le débat public ignore.

Rupture dans la longue histoire du capitalisme industriel

L’IA générative constitue une rupture dans cette longue histoire. Pour la première fois depuis l’industrialisation, les tâches cognitives qualifiées, rédaction, analyse documentaire, synthèse, production de premiers jets, se retrouvent directement exposées. Eloundou, Manning, Mishkin et Rock estiment qu’environ 80 % de la population active états-unienne pourrait voir au moins 10 % de ses tâches affectées par les grands modèles de langage, et que cette exposition croît avec le niveau de salaire. C’est l’exact inverse du schéma observé lors de toutes les vagues précédentes.

Le cadre analytique développé par Acemoglu et Restrepo permet d’aller plus loin. Leur modèle distingue deux effets opposés produits par toute vague d’automatisation :

• l’effet de déplacement, d’abord : des travailleurs perdent des tâches au bénéfice de la machine, ce qui réduit mécaniquement la demande de travail et pèse sur les salaires des groupes affectés ;

• l’effet de réintégration, ensuite : l’automatisation produit de nouvelles tâches où la valeur humaine est décisive, générant une demande compensatrice.

L’histoire longue du capitalisme industriel peut se lire comme une succession de ces deux effets, le second finissant généralement par compenser le premier.

Le cas de la traduction permet de voir très concrètement comment déplacement et réintégration se combinent, l’IA générative peut produire en quelques secondes un premier jet dans une autre langue, ce qui déplace une partie du travail auparavant effectué par des traducteurs humains vers la machine. Mais cette automatisation réintègre simultanément d’autres tâches ou renforce leur importance, telles que la vérification des contresens, l’adaptation au contexte culturel, l’harmonisation de la terminologie, le contrôle de la qualité et la validation finale.

Potentiel déséquilibre

Ce qui est préoccupant avec l’IA générative, c’est le déséquilibre potentiel entre ces deux dynamiques. Le déplacement s’opère à une vitesse que les marchés du travail et les institutions de formation peinent à absorber, tandis que la réintégration reste encore largement à construire.

Cependant, le phénomène le plus important n’est pas sectoriel, mais il est interne aux métiers eux-mêmes. Dans ses « Perspectives de l’emploi », l’OCDE met en évidence que les professions les plus exposées à l’IA générative sont précisément celles à forte densité cognitive : finance, droit, conseil, enseignement supérieur. Contrairement aux vagues précédentes qui frappaient les zones rurales et les bassins industriels, l’exposition est désormais plus forte dans les grandes métropoles et chez les travailleurs hautement qualifiés, un renversement géographique et social inédit.

Redistribuer les tâches

Ce renversement se joue concrètement au niveau de la tâche.

Dans un même poste d’analyste financier ou de juriste, certaines tâches migrent vers l’IA (produire un résumé exécutif, générer une première analyse de contrat, synthétiser une revue de littérature), tandis que d’autres se revalorisent mécaniquement : définir le cadre d’analyse pertinent, évaluer la qualité d’un raisonnement automatisé, détecter une erreur factuelle dans un output, assumer la responsabilité juridique ou éthique d’une décision. Ce ne sont pas des emplois qui disparaissent. Ce sont des bouquets de tâches qui se redistribuent entre humains et machines, transformant de l’intérieur ce qu’un employeur attend d’un salarié qualifié.

Cette redistribution des tâches a une implication directe sur les compétences qui seront réellement valorisées dans les années à venir, et elle renverse une partie des évidences habituelles sur la formation professionnelle.

Former les travailleurs à utiliser l’IA au sens instrumental, maîtriser un outil, rédiger des prompts efficaces, s’approprier une interface, est utile à court terme, mais c’est insuffisant si la compétence réellement demandée demain n’est pas de produire avec l’IA, mais de superviser et de critiquer ce qu’elle produit.

Un enjeu de formation

Or, superviser efficacement un output d’IA requiert exactement ce que les formations courtes et techniques peinent à développer : une culture générale solide permettant de détecter une erreur de fond, une capacité argumentative pour évaluer la cohérence d’un raisonnement, une connaissance des biais cognitifs pour identifier les angles morts d’une analyse automatisée. Ce sont des compétences que les sciences de l’éducation regroupent sous le terme de métacompétences : apprendre à apprendre, à exercer un jugement critique, à mobiliser des savoirs dans des situations inédites.

Le paradoxe devient alors le suivant. À mesure que l’IA automatise les tâches routinières de la connaissance, elle valorise précisément ce que les formations généralistes et les cursus de sciences humaines cultivent de longue date et que les débats sur l’employabilité ont eu tendance à déconsidérer au profit de compétences techniques plus immédiatement mesurables.

Non par nostalgie des humanités, mais par logique économique pure. Si la machine produit le texte, l’analyse et la synthèse, la valeur marginale de l’humain réside dans sa capacité à juger si ce texte dit vrai, si cette analyse est pertinente au regard du contexte réel, si cette synthèse sert l’objectif poursuivi.

Hugo Spring-Ragain ne travaille pas, ne conseille pas, ne possède pas de parts, ne reçoit pas de fonds d’une organisation qui pourrait tirer profit de cet article, et n’a déclaré aucune autre affiliation que son organisme de recherche.

– ref. L’IA générative ne détruira pas votre emploi mais elle va changer profondément votre métier – https://theconversation.com/lia-generative-ne-detruira-pas-votre-emploi-mais-elle-va-changer-profondement-votre-metier-279911

Source: MIL-OSI-Submissions-French

Source: The Conversation – France (in French) – By Hugo Spring-Ragain, Doctorant en économie / économie mathématique, Centre d’études diplomatiques et stratégiques (CEDS)

L’intelligence artificielle ne détruit pas tant des emplois qu’elle modifie profondément les compétences nécessaires pour les accomplir. De cette confusion entre emploi et compétences risquent de naître des erreurs dans les politiques d’accompagnement des mutations en cours.

---

Chaque grande vague technologique a produit son lot de prédictions contradictoires sur l’emploi. L’intelligence artificielle (IA) ne fait pas exception. Mais avant de savoir combien d’emplois l’IA va créer ou détruire, il faudrait s’accorder sur ce qu’elle automatise réellement. La réponse oblige à distinguer trois notions que le débat public confond régulièrement : l’emploi, la compétence et la tâche.

Les grandes vagues d’automatisation ont suivi une logique remarquablement stable en deux siècles : vapeur, électricité, robotique industrielle ont déplacé les tâches physiques répétitives et épargné le travail cognitif non routinier. Cette régularité empirique a été formalisée par Autor, Levy et Murnane dès 2003 sous le nom d’« hypothèse de polarisation des tâches ».

Une illusion persistante

L’automatisation ronge les emplois intermédiaires, ceux des cols bleus qualifiés et employés de bureau exécutant des tâches routinières, mais épargne les deux extrémités. D’un côté, les tâches manuelles non routinières, comme la plomberie ou les soins, de l’autre, les tâches cognitives non routinières, comme l’analyse, le conseil ou la rédaction experte. Ces dernières constituaient le cœur des professions du tertiaire qualifié, et la conviction s’était solidement installée qu’elles resteraient hors d’atteinte.

---

À lire aussi :
Pourquoi l’IA oblige les entreprises à repenser la valeur du travail

---

Cette conviction reposait sur une confusion conceptuelle qu’il faut dissiper avant tout. Ce n’est pas l’emploi de juriste ou d’analyste financier qui était protégé, c’est un ensemble de tâches précises qui composaient cet emploi et qui résistaient jusqu’ici à l’automatisation. La distinction entre ces trois niveaux est fondamentale.

Un emploi désigne un poste occupé dans une organisation, avec un contrat, un salaire, une fiche de poste. Une compétence est une capacité cognitive ou technique mobilisable dans plusieurs contextes professionnels. Une tâche est une action précise, délimitable, dont on peut évaluer si elle est ou non automatisable à un coût donné. C’est à ce troisième niveau que se joue réellement la transformation en cours, et c’est précisément ce niveau que le débat public ignore.

Rupture dans la longue histoire du capitalisme industriel

L’IA générative constitue une rupture dans cette longue histoire. Pour la première fois depuis l’industrialisation, les tâches cognitives qualifiées, rédaction, analyse documentaire, synthèse, production de premiers jets, se retrouvent directement exposées. Eloundou, Manning, Mishkin et Rock estiment qu’environ 80 % de la population active états-unienne pourrait voir au moins 10 % de ses tâches affectées par les grands modèles de langage, et que cette exposition croît avec le niveau de salaire. C’est l’exact inverse du schéma observé lors de toutes les vagues précédentes.

Le cadre analytique développé par Acemoglu et Restrepo permet d’aller plus loin. Leur modèle distingue deux effets opposés produits par toute vague d’automatisation :

• l’effet de déplacement, d’abord : des travailleurs perdent des tâches au bénéfice de la machine, ce qui réduit mécaniquement la demande de travail et pèse sur les salaires des groupes affectés ;

• l’effet de réintégration, ensuite : l’automatisation produit de nouvelles tâches où la valeur humaine est décisive, générant une demande compensatrice.

L’histoire longue du capitalisme industriel peut se lire comme une succession de ces deux effets, le second finissant généralement par compenser le premier.

Le cas de la traduction permet de voir très concrètement comment déplacement et réintégration se combinent, l’IA générative peut produire en quelques secondes un premier jet dans une autre langue, ce qui déplace une partie du travail auparavant effectué par des traducteurs humains vers la machine. Mais cette automatisation réintègre simultanément d’autres tâches ou renforce leur importance, telles que la vérification des contresens, l’adaptation au contexte culturel, l’harmonisation de la terminologie, le contrôle de la qualité et la validation finale.

Potentiel déséquilibre

Ce qui est préoccupant avec l’IA générative, c’est le déséquilibre potentiel entre ces deux dynamiques. Le déplacement s’opère à une vitesse que les marchés du travail et les institutions de formation peinent à absorber, tandis que la réintégration reste encore largement à construire.

Cependant, le phénomène le plus important n’est pas sectoriel, mais il est interne aux métiers eux-mêmes. Dans ses « Perspectives de l’emploi », l’OCDE met en évidence que les professions les plus exposées à l’IA générative sont précisément celles à forte densité cognitive : finance, droit, conseil, enseignement supérieur. Contrairement aux vagues précédentes qui frappaient les zones rurales et les bassins industriels, l’exposition est désormais plus forte dans les grandes métropoles et chez les travailleurs hautement qualifiés, un renversement géographique et social inédit.

Redistribuer les tâches

Ce renversement se joue concrètement au niveau de la tâche.

Dans un même poste d’analyste financier ou de juriste, certaines tâches migrent vers l’IA (produire un résumé exécutif, générer une première analyse de contrat, synthétiser une revue de littérature), tandis que d’autres se revalorisent mécaniquement : définir le cadre d’analyse pertinent, évaluer la qualité d’un raisonnement automatisé, détecter une erreur factuelle dans un output, assumer la responsabilité juridique ou éthique d’une décision. Ce ne sont pas des emplois qui disparaissent. Ce sont des bouquets de tâches qui se redistribuent entre humains et machines, transformant de l’intérieur ce qu’un employeur attend d’un salarié qualifié.

Cette redistribution des tâches a une implication directe sur les compétences qui seront réellement valorisées dans les années à venir, et elle renverse une partie des évidences habituelles sur la formation professionnelle.

Former les travailleurs à utiliser l’IA au sens instrumental, maîtriser un outil, rédiger des prompts efficaces, s’approprier une interface, est utile à court terme, mais c’est insuffisant si la compétence réellement demandée demain n’est pas de produire avec l’IA, mais de superviser et de critiquer ce qu’elle produit.

Un enjeu de formation

Or, superviser efficacement un output d’IA requiert exactement ce que les formations courtes et techniques peinent à développer : une culture générale solide permettant de détecter une erreur de fond, une capacité argumentative pour évaluer la cohérence d’un raisonnement, une connaissance des biais cognitifs pour identifier les angles morts d’une analyse automatisée. Ce sont des compétences que les sciences de l’éducation regroupent sous le terme de métacompétences : apprendre à apprendre, à exercer un jugement critique, à mobiliser des savoirs dans des situations inédites.

Le paradoxe devient alors le suivant. À mesure que l’IA automatise les tâches routinières de la connaissance, elle valorise précisément ce que les formations généralistes et les cursus de sciences humaines cultivent de longue date et que les débats sur l’employabilité ont eu tendance à déconsidérer au profit de compétences techniques plus immédiatement mesurables.

Non par nostalgie des humanités, mais par logique économique pure. Si la machine produit le texte, l’analyse et la synthèse, la valeur marginale de l’humain réside dans sa capacité à juger si ce texte dit vrai, si cette analyse est pertinente au regard du contexte réel, si cette synthèse sert l’objectif poursuivi.

Hugo Spring-Ragain ne travaille pas, ne conseille pas, ne possède pas de parts, ne reçoit pas de fonds d’une organisation qui pourrait tirer profit de cet article, et n’a déclaré aucune autre affiliation que son organisme de recherche.

– ref. L’IA générative ne détruira pas votre emploi mais elle va le changer profondément votre métier – https://theconversation.com/lia-generative-ne-detruira-pas-votre-emploi-mais-elle-va-le-changer-profondement-votre-metier-279911