LONDON — Prime Minister Andy Burnham’s government said the U.K. is willing to discuss American concerns over its digital services tax amid renewed pressure from the White House.

President Donald Trump in June threatened to impose 100 percent tariffs on European countries with DSTs which target U.S. tech firms, and in an interview with The Times newspaper published on Monday, Trump’s top trade official Jamieson Greer said the threat was “not a bluff” and the president’s demands that foreign governments abandon such taxes were “quite serious.”

“We remain open to discussing U.S. concerns and working with partners internationally,” a U.K. government spokesperson said when asked about Greer’s comments.

“This tax is about making sure that businesses pay their fair share of U.K. tax based on the value they derive from U.K. activities,” the spokesperson said, adding that the U.K. is committed to removing it “once a global solution is in place.”

The DST raised over £1 billion last year, predominantly from American tech firms, and has repeatedly drawn Trump’s ire.

The U.K. government has so far resisted calls to abandon the tax, including during trade negotiations last year. A U.K.-U.S. Economic Prosperity Deal signed by Trump and Burnham’s predecessor, Keir Starmer, did not mention the DST but said both sides would continue discussions to increase digital trade and address non-tariff barriers.

Greer added in his interview with The Times that the U.S. administration would not “set artificial timelines” and that relations with his British counterparts, including Trade Secretary Jonathan Reynolds and the prime minister’s business adviser Varun Chandra, are positive.

A bipartisan House delegation met with top Vatican officials on Wednesday to discuss artificial intelligence during a visit that briefly included Pope Leo XIV, according to two people with knowledge of the trip granted anonymity to disclose details of the private meetings.

The delegation’s visit, which has not been previously reported, included nine lawmakers and was coordinated by the House Select Committee on the Chinese Communist Party, the person said. Committee member Rep. Dan Newhouse (R-Wash.) and California Rep. Ro Khanna, the panel’s top Democrat, led the trip.

A spokesperson for the committee declined to comment.

In an interview confirming the visit, Khanna said the topics of discussion included religious freedom, AI’s economic and geopolitical risks, as well as broader ethical and existential implications. It centered on two key principles: human dignity and equality.

“Human dignity means that human beings need to be responsible for any key decisions involving health, involving finances, involving individual freedom, involving public services,” he said.

The Silicon Valley Democrat said that in this context, equality means preventing automation from causing mass unemployment, and ensuring AI cannot be used to exclude people from social services. Khanna said it also means guarding against a concentration of power in which “a few billionaires can make decisions about data and algorithms.”

The Vatican did not respond to a request for comment about the meeting. But Pope Leo has taken a special interest in the technology and used a May encyclical to call on countries to “safeguard humanity” as AI develops, warning that the technology could deepen inequality, fray the social fabric and erode moral responsibility if it doesn’t have ethical guardrails. Congress, meanwhile, has introduced dozens of AI bills and held several hearings, but has failed to agree on a comprehensive federal framework for governing the technology.

The pope previously cast the global race to develop cutting-edge AI technology in stark terms. As the U.S., China and other powers compete for an edge, he warned against a “dehumanizing ambition to develop ever more powerful technologies or to secure control over them,” describing a contest between “opposing imperialisms” seeking either to preserve or seize technological supremacy.

Khanna told POLITICO he plans to introduce legislation that would tax the use of autonomous AI agents to incentivize companies to hire human workers.

The group also discussed AI’s implications for human mortality and the limits of human knowledge — boundaries some developers have long imagined technology might overcome by uploading human consciousness and accumulated wisdom to the cloud.

Following the meetings, Khanna sent letters to top executives at OpenAI, Anthropic, Meta, Google, Microsoft, Apple, Nvidia, and Amazon, asking how they intend to align their technology with the tenets of human dignity and equality.

An OpenAI spokesperson pointed POLITICO to prior policy documents outlining the AI company’s commitment to safety and to ensuring benefits from the technology are distributed equitably. The other companies did not respond to requests for comment.

The lawmakers traveled to Berlin to discuss China’s rising global influence, before picking up on the topic during their Vatican visit and in meetings with Italian government officials a few days later.

Alongside Khanna and Newhouse, Reps. Diana DeGette (D-Colo.), Buddy Carter (R-Ga.), John Rutherford (R-Fla.), Randy Feenstra (R-Iowa), Shontel Brown (D-Ohio), Jill Tokuda (D-Hawaii) and Raja Krishnamoorthi (D-Ill.) took part in the discussions, according to one of the people familiar with the trip.

Spokespeople for the members did not respond to requests for comment.

NATO plans to fortify its eastern edge using drones, sensors and artificial intelligence in a massive network designed to spot and help repel a potential Russian attack.

Western officials say the alliance is exploring AI-assisted drone operations as part of the broad effort to bolster its defenses. The aim, however, isn’t full autonomy; instead, machines could help to pilot the drones while human operators make the call on what they strike.

The approach echoes what Ukraine is already doing. The embattled country is increasingly incorporating AI into its drone operations, but both soldiers and defense tech firms have repeatedly emphasized that humans must remain at the center of the fight, even as the war becomes more robotic.

For NATO, AI-assisted drones would be part of the Eastern Flank Deterrence Initiative — a digital battlespace network the alliance is building to track threats along its borders with Russia and Belarus, which stretch from northern Finland to the Black Sea.

The new concept includes thousands of drones, sensors and satellites linked with AI to detect potential breaches of allied territory and repel attacks as the first line of defense. Conventional military forces, such as tanks, fighter jets and artillery, will remain the backbone.

U.S. Army Maj. Ben Schiff, a software operations officer, told Bild — which, like Business Insider and POLITICO, is part of the Axel Springer Global Reporters Network — that both Ukraine and Russia still rely on humans to control drones.

While some systems have AI or machine-learning capabilities that allow brief moments of autonomy in flight, piloting is primarily performed by operators on the ground. Schiff said that “the most boring parts of flying”— drones loitering above the battlefield looking for targets — should be automated. But when it comes to deciding what the drone should strike, that’s where a human comes into play.

“So the humans are deciding what the drone does, but they don’t necessarily need to take the action of flying it because we don’t have a million pilots. We can’t fly a million drones at the same time,” said Schiff at the U.S. Army Europe and Africa headquarters in Wiesbaden, Germany.

Schiff is a product manager for the EFDI Data Backbone, a developing framework designed to help NATO forces detect, classify, share and act on information at speed and scale. The backbone bridges incompatible computer and command systems across multiple allied nations, allowing real-time data to flow from forward-deployed sensors all the way back to headquarters.

The goal is to help NATO move beyond today’s linear system — called a “kill chain” — where information moves from the front line back to a decision-maker and then back again to the battlefront to a “kill web,” where data from sensors can be rapidly passed to weapons systems and operators.

“The value of a drone, sensor or other capability is not determined solely by its individual performance,” said Maj. Matt Blubaugh, a spokesperson for U.S. Army Europe and Africa. “Its effectiveness depends on how well it integrates into the broader operational ecosystem.”

Learning from Ukraine

Ukraine is already pioneering the integration of AI into military operations.

On the battlefield, Ukraine has made significant progress in using AI to support functions like decision-making, drone navigation, terminal guidance and precision, said Andrii Hrytseniuk, the CEO of the Ukrainian state-backed innovation platform Brave1.

This spring, Hrytseniuk told Business Insider in Kyiv that more than 200 local companies were developing AI-based components for drones, missiles, radars and other systems. However, they insist machines won’t control drones from start to finish.

Wild Hornets, the manufacturer of the popular Sting interceptor drone, also told Business Insider in May that AI will be gradually integrated at almost every stage of flight operations — target detection, identification and, later, terminal guidance.

A spokesperson for the company said humans will still make the final decision on what to strike. Since Russia is constantly changing its tactics, interceptor operators must keep tabs on these changes and make necessary adjustments, they said.

As autonomous systems become more common on the battlefield in Ukraine, commanders also told Business Insider that humans need to remain in the loop.

Grek, the call sign of a company commander overseeing robots with Ukraine’s 21st Unmanned Systems “Kraken” Regiment, said the goal is to “allow people to focus on the tasks where human judgment has the greatest value.”

U.S. Army officers stressed that NATO is using the operational environment in Ukraine to inform military planning for a potential future confrontation.

“We’re not copying Ukraine,” said Capt. Ronan Sefton. “We’re learning from Ukraine.”

The Axel Springer Global Reporters Network harnesses the resources of the company’s newsrooms to publish ambitious scoops, investigations, interviews, opinion pieces and analysis. It allows journalists — including those from POLITICO, Business Insider, WELT, BILD, The Telegraph, Onet and Fakt — to collaborate on major stories for an international audience of hundreds of millions across platforms: online, print, TV and audio.

President Donald Trump is paving a legal pathway for U.S. companies to launch cyberattacks on foreign cybercriminal gangs — a significant and potentially controversial measure that would put approved tech and cybersecurity firms on the front lines of digital combat.

The presidential memorandum, released late Wednesday, comes as the Trump administration has repeatedly pushed for more aggressive action to counter foreign scams and cyberattacks, which the White House said cost Americans nearly $21 billion last year.

The memo represents one of the biggest shifts in U.S. cyber policy undertaken in recent years. It would empower tech and security companies — whose data and control over internet infrastructure often offer unique insight into foreign hacking operations — to mount state-sanctioned digital strikes.

While many such companies already work closely with U.S. intelligence and law enforcement agencies, a web of legal and political constraints has long prevented them from taking direct action inside foreign networks.

Companies that want to participate would be required to sign contracts with both the Department of Justice and the Department of Homeland Security and to undergo what the memo describes as “rigorous vetting” while working with the government. The overall effort would be overseen by a National Coordination Center, established in an earlier Trump administration executive order, with co-executive directors from DOJ and DHS.

However, the memo states that no operations by the companies would be approved until the executive directors at DOJ and DHS establish “consensus procedures” with the White House Homeland Security Council guaranteeing “complete oversight and control of Participating Companies’ performance.”

Those procedures, it notes, should be drafted within 60 days. They are likely to be extensive.

They will outline steps for participating companies to obtain approval for proposed offensive hacking operations, so the government can confirm that the targets are criminal gangs and ensure that operations are consistent with U.S. law and don’t undermine ongoing U.S. intelligence efforts. Companies could propose surveillance operations to help identify criminals or “effects” operations to degrade the systems they use to stage their attacks.

Participating companies would have to pass minimum standards for technical expertise and personnel vetting, and would be required to notify the federal government if they believe approved operations may result in the loss of life or rise to the level of use of force under international law.

Some see the memo as a critical step to help the U.S. government counter foreign cybercriminal gangs that operate outside the reach of U.S. law enforcement.

“For years we’ve called the American technology industry a strategic asset but left it on the cyber sidelines,” Joe Lin, the CEO and co-founder of Twenty, a start-up that builds offensive cyber tools for the U.S. government, said in a statement. “This administration is changing the paradigm.”

The memo notes that companies will only be authorized to target criminals that are “not an institutional part of a foreign government or wholly operated under a foreign government’s direction.”

Even with the help of the U.S. intelligence community, making that distinction could be difficult.

Adversaries such as Russia, China and Iran have persistently targeted U.S. critical infrastructure, including water systems, ports, and telecommunications infrastructure, while multinational crime syndicates have defrauded billions of dollars annually from Americans via complex online schemes.

But many cyber gangs in Eastern Europe are thought to operate with the tacit consent of the Russian government, while state hackers in Iran and China sometimes moonlight as cybercriminals to earn extra money or deflect blame for their governments’ attacks.

More broadly, it is not always easy for digital investigators to determine who is responsible for a given cyberattack, or who different computer networks belong to — another risk the memo contemplates.

Companies that accidentally carry out operations targeting a U.S. citizen or network will be required to immediately pause the operation and notify the U.S. government, the memo states. It does not appear to preclude activities that are deliberately “directed” at a U.S. person, so long as they receive “any necessary authorization, judicial or otherwise, prior to approval of the operation.” Under U.S. law, a “U.S. person” can refer to an American business or organization.

Many lawmakers and security experts have broadly supported calls for the private sector to play a larger role in responding to cybercrime, though not all approve of granting them the ability to launch active hacking efforts.

In recent years, some House members have debated the idea of issuing “letters of marque” to private companies to carry out cyberattacks on behalf of the U.S. government, similar to the U.S. Navy authorizing private ships to disrupt British shipping during the War of 1812.

As part of a more assertive cyber posture, Trump has turned to U.S. Cyber Command to mount digital attacks in tandem with U.S. military operations, including in Iranand Venezuela. He signed an executive order this March to clamp down on countries that fail to take action against scam centers operating within their borders.

That same month, the White House called on the private sector to broadly help it “disrupt” foreign adversaries in its new national cyber strategy, though it stopped short of telling private companies to take riskier and more consequential steps, such as directly launching attacks against foreign criminals.

Some of the most prolific online fraud operations are believed to emanate from scam compounds in Southeast Asia. But hackers from North Korea — who for years have stolen hundreds of millions in cryptocurrency from victims around the world — would likely be exempt from targeting by U.S. companies since they work at the direction of the North Korean government.

A news organization and free speech nonprofit sued President Donald Trump over his social media platform’s program that offers paid early access to his posts after the scheme went live at the beginning of August.

The lawsuit, filed Wednesday in the U.S. District Court for the Southern District of New York by the left-leaning nonprofit newsroom The Intercept and the Freedom of the Press Foundation, alleges the practice “is extraordinary, corrupt, and unconstitutional.”

The plaintiffs also say the controversial $100,000-a-month program violates their First Amendment rights to access presidential statements “on equal terms with other members of the press and public.”

In announcing plans to introduce the program, Truth Social — whose parent company Trump founded in 2021 after being kicked off social media platforms and owns a plurality stake in — described the initiative as valuable for “organizations that place a premium on immediate, verified access to information.” But critics allege it amounts to insider trading on advanced access to the president’s market-moving words.

The service gives traders and other entities high-speed access to the president’s Truth Social account in addition to the nine other most-popular accounts on the platform, including Vice President JD Vance, White House press secretary Karoline Leavitt and Health and Human Services Secretary Robert F. Kennedy Jr.

A spokesperson for Truth Social, which isn’t named as a defendant in the complaint but is referenced throughout it, said in a statement that “information from President Trump is disseminated by countless platforms and news outlets, many of which offer subscription APIs.”

“One of those channels is Truth Social, which was founded as an uncancellable haven for free speech after the President was unjustly deplatformed,” the statement continued. “Now, left-wing activists are trying to wrongfully weaponize the courts to censor him again and harm our shareholders.”

The White House did not respond to requests for comment.

The Intercept’s chief legal officer, David Bralow, in a statement said “nothing could be more antithetical to the free, independent press than the president charging for early access to his public announcement.”

Natalie Harp, an executive assistant to Trump who media reports suggest is responsible for authoring many of the president’s social media posts, is named as a defendant in the suit alongside deputy chief of staff Dan Scavino, the Executive Office of the President and the White House Office.

The subjects of the president’s Truth Social posts can range from squabbles with federal judges to hiring and firings within his administration and threats against foreign adversaries. They also provide a glimpse at the issues of the day that occupy Trump’s attention.

Such announcements have had the ability to sway markets, like when oil prices plunged after Trump posted that he was calling off an attack on Iran in April. He has also occasionally touted specific companies on the platform, causing their stock prices to rise.

The plaintiffs also express concerns that the program would give outlets willing to pay for the service an unfair advantage over other newsrooms and hinder efforts to catalog the president’s posts by scraping Truth Social.

The company said in an August earnings report that despite a $238 million loss in the second quarter of 2026, 10 customers had already registered for the service.

The program has caught the eyes of Democratic lawmakers on the Hill. Sens. Ruben Gallego (D-Ariz.) and Mark Warner (D-Va.) introduced a bill Tuesday seeking to ban social media companies from selling early access to government employees’ accounts and specifically invoked the Truth Social program.

LAS VEGAS — Weeks before they escaped a closed test and launched a cyberattack without any human prompting, some of OpenAI’s most advanced artificial intelligence agents secretly began sharing tips on how to cheat their way through an internal hacking evaluation, two of the AI company’s researchers said Wednesday.

“This is a pivotal moment both for our company as well as the AI industry as a whole,” Michael Dalton, one of the AI-maker’s researchers, said during a presentation at the Black Hat cybersecurity conference in Las Vegas, Nevada.

The latest disclosure provides greater detail on the timeline and methods used by two of OpenAI’s models before they slipped outside a controlled environment and onto the open internet, allowing the models to breach AI developer platform Hugging Face undetected. OpenAI admitted its models were responsible for the hack late last month, roughly a week after Hugging Face said an autonomous AI system broke into its network.

The revelation is likely to spur fresh scrutiny of the measures leading AI labs such as OpenAI and Anthropic are taking to monitor the behavior of their most cyber-capable technology — especially during evaluations where agents are prompted to demonstrate their hacking skills in what is meant to be a controlled setting.

On Tuesday, the U.K.’s AI Safety and Security Institute disclosed that Anthropic’s most powerful AI model created fake online personas and sought to trick a human coder into abetting a cyberattack during a recent hacking test gone wrong. After the Hugging Face disclosure last month, Anthropic conducted a review and found models it was testing had breached three organizations in separate incidents dating back to April.

Dalton and Eric Wallace, another OpenAI researcher, said Wednesday the AI giant recently learned that multiple agents it was testing simultaneously began communicating over an internal message board in early May. There, different models shared advice about how to accomplish difficult hacking challenges they were struggling to surmount, including workarounds that required internet access.

Two OpenAI models ultimately strung together a series of sophisticated techniques to gain access to the internet and worm their way inside Hugging Face in mid-July. OpenAI has said the models were focused on completing a hacking evaluation they were prompted to solve, and that correct answers could be found on the AI developer platform.

The OpenAI researchers told conference attendees that since early May, the models created a message board inside OpenAI’s Artifactory internal file system. Without the company’s knowledge, the models spent months independently exchanging information and techniques to help each other complete difficult tasks.

Wallace said that when models get stuck, they often “try to game or cheat the task in order to get their reward.”

“The beginning of this incident kind of traces back to this point where one agent realized it could communicate by uploading a note,” he added.

By late May, one model found a way to abuse Artifactory’s internet access to retrieve files from various websites — effectively giving the models indirect access to the internet. In June, the models were able to find and exploit a previously unknown zero-day vulnerability to gain administrative privileges on Artifactory, install remote software and expand their activity within the system.

These efforts eventually overloaded Artifactory, causing a system-wide outage in early July that alerted OpenAI’s engineers to the ploy. After investigating, the company revoked the model’s credentials, removed the message board and worked with Artifactory to fix any gaps before resuming training. But the models found another way to communicate inside Artifactory just days later and continued exchanging techniques to target additional vulnerabilities within OpenAI’s infrastructure and external systems, including Hugging Face.

In light of the incident, Dalton said OpenAI is “consciously slowing down research to enhance security and to upgrade the security principles and foundation of our environment, and dramatically scaling up the monitoring of our AI agents and improving our general security control environment across prevention, detection, and mitigation.”

Brussels is bracing for an unusual fight between the EU’s smallest country and a British ex-footballing legend.

Peter Shilton, the England goalkeeper who conceded the “Hand of God” goal from Diego Armando Maradona in 1986, has started a new life as an anti-gambling advocate after overcoming a decades-long addiction.

Despite being a diehard Brexit supporter, he’s become the poster boy of the European Parliament’s push to tax online betting in a bid to raise some much-needed funds to finance the bloc’s next €2 trillion budget.

But the campaign has run into strong opposition from Malta. The tiny island in the Mediterranean Sea, with a population of just over half a million people, is home to a burgeoning betting sector. It says that higher taxes will cripple its gambling industry, boost illegal operators and drive firms outside the bloc.

“[Malta] will not accept the introduction of any EU-level taxes designed to sustain the bloc’s spending,” the country’s Prime Minister, Robert Abela, told the Maltese Parliament in June.

But Shilton, who lost more than £1 million in betting on horse racing over 45 years and now runs his own gambling addiction charity, dismisses the arguments by Malta and the gambling lobbies as “window dressing.” He’s in favor of higher taxes as he wants to shrink advertising revenue that is used to lure in new gamblers.

“Deep down they’re after everybody’s money. Simple as that,” he told POLITICO during a visit to Brussels in June.

The topic has split the EU’s 27 governments, pitting gambling-heavy Southern European countries against their more supportive Western European peers, led by France. Capitals are already fighting even though the Commission hasn’t yet issued a formal proposal for the possible tax, which would ultimately need to be unanimously approved by governments.

It’s one of numerous budget battle lines being drawn, with Ireland — which is steering the talks as chair of the rotating Council presidency — set to restart negotiations to facilitate an overall deal on the EU budget before the end of the year.

That’s no mean feat given Dublin’s task to mesh competing spending priorities into a single budget — financing everything from farmers’ subsidies to foreign aid — that is acceptable for each of the EU’s 27 governments.

National capitals will have to unanimously approve new EU-wide taxes — known as own resources — to pay for soaring defense spending and post-Covid debt repayments if they want to avoid drastically increasing national contributions to Brussels.

Supporters of the gambling levy point to the fact that it would rake in over €13 billion throughout the next budget cycle and — for some, more importantly — address a serious public health issue. An estimated 80 million adults globally have experienced a gambling addiction, according to experts.

“We look on it [gambling] as an illness. It’s something that’s inborn in you and that can be ignited,” Shilton said.

Malta’s game plan

Malta has invested heavily in the gambling industry — including lotteries, betting and casinos increasingly operating online — which now accounts for around 12 percent of its gross domestic product.

These firms have relocated to Malta because of its light-touch licensing regime, business-friendly tax regime and balmy weather.

The country is “as dependent on the online gambling industry as Germany is on cars,” said an EU diplomat, granted anonymity to speak freely.

While gambling firms need local authorization to operate in most other European countries, securing the Maltese license is crucial to access banking services and gain a foothold in the EU market.

Malta-based firms dominated the German and Austrian online gambling markets before national regulators cracked down. This has prompted the Maltese government to refuse to recognize some court rulings and sanctions issued by other EU countries against its gambling firms.

Given its influence, it is hardly surprising that the gambling industry has found a friendly ear among Malta’s politicians in Brussels.

The Maltese president of the European Parliament, Roberta Metsola, last year gave the opening speech at an international gambling conference in Rome that also featured Italian Foreign Affairs Minister Antonio Tajani.

“I’m more than a little proud that it started in my island home of Malta,” she said, referring to SiGMA, a Maltese events company that focuses on online gambling founded by Eman Pulis, a university friend of Metsola.

Betting lobbies say they oppose higher gambling rates on the grounds that they will fuel appetite for the illegal market, away from the grasp of EU rules.

“A higher tax would lead to worse odds for the customers … and it is relevant because access to the illegal markets in Europe is, obviously, one click away,” said secretary general of the European Gaming and Betting Association, Maarten Haijer.

Nicola Matteucci, an economist at the Università Politecnica delle Marche in Italy who has undertaken extensive research on the gambling sector, argued there is a “point where prices exceed a certain level and the demand [for gambling] diminishes. But it’s not as immediate as suggested by the industry.”

Matteucci said that most gamblers will be undeterred by slightly higher taxes and worse odds as they are not fully rational consumers.

Anti-gambling groups reason instead that higher taxes will reduce the sector’s spending on commercials, preventing would-be punters from getting sucked in to gambling in the first place.

“Higher taxes will therefore mean less gambling advertising overall and many people would regard that as a public benefit,” said Derek Webb, the founder of the Campaign for Fairer Gambling advocacy group.

Club Med joins Malta

Malta has joined forces with fellow Mediterranean countries — Italy, Portugal and Spain — to challenge the mooted tax which was first proposed by the Parliament’s socialist lawmaker Victor Negrescu, said four diplomats with knowledge of the discussions.

According to the European Commission’s estimates, seen by POLITICO, a 3 percent tax on the net turnover of the online gambling sector would generate an estimated €1.9 billion per year.

With its big online gambling market, Spain is expected to be among the biggest financial losers, should the tax go ahead. It is estimated to be on the hook for €414 million per year, almost a quarter of the total amount. That compares to a projected bill of €165 million per year for Malta— a disproportionality high amount for such a small country.

Portugal is also reluctant to back the levy. It fears that higher taxes would eat into revenue brought in by state-run betting and lotteries that is currently channeled to the charity Santa Casa da Misericórdia de Lisboa‘s healthcare and youth support programs, said a Portuguese official.

Meanwhile, given the relatively low uptake of online gambling, Italy’s misgivings have surprised anti-betting advocates. Rome is expected to pay a mere 7 percent of the proposed new levy — a significantly lower proportion than its regular EU budget contributions.

However, Prime Minister Giorgia Meloni’s Brothers of Italy party has previously been receptive to the gambling industry. Last year its MPs passed a resolution encouraging the reversal of a ban on professional football clubs advertising gambling firms.  

LONDON — The U.K. capital’s transport authority has granted approval for Wayve and Uber to begin giving rides to members of the public in autonomous vehicles.

Transport for London (TfL) said it licensed 15 modified vehicles operated by the companies, which have a partnership, as “Private Hire Vehicles” (PHV) on a trial basis.

In a statement, London-based startup Wayve said the licenses were “an important step forward” that will allow it to begin giving rides to a small number of passengers later this summer ahead of a full public launch.

Wayve said its vehicles “are designed to operate autonomously, and will do the driving,” though under TfL’s rules, a licensed PHV driver must be present and responsible for the vehicle at all times.

“Safety is our top priority,“ a TfL spokesperson said. “Any new vehicle licensed to carry passengers on London’s roads must align with our Vision Zero goal of eliminating all deaths and serious injuries from collisions on London’s streets by 2041.”

Successive U.K. governments have sought to make the country a European pioneer in self-driving technology.

The Department for Transport opened a permitting scheme for companies to operate commercial robotaxi services without a human driver in May. Applications for that scheme — which is separate from TfL’s PHV regime — continue to be assessed by central government with input from local transport authorities including TfL.

Getting licenses isn’t the only obstacle facing robotaxi services. A survey by the London Assembly’s Transport Committee this month identified widespread opposition to autonomous passenger vehicles among the capital’s inhabitants, with just 29 percent of Londoners saying they support the roll out.

Leading artificial intelligence models from Anthropic and OpenAI created fake online personas and tried to deceive human coders into abetting a cyberattack during a recent safety evaluation, the U.K.’s AI Safety and Security Institute disclosed Tuesday.

It marks the latest case in which a powerful AI system has attempted a digital attack on an unwitting third party without direct prompting during such an evaluation — heightening concerns the powerful technology is advancing too fast for responsible oversight.

The disclosure is likely to ignite fresh calls in Washington and Silicon Valley for more rigorous regulation of the AI industry, particularly over frontier models with advanced capabilities to detect and launch cyberattacks. It comes just days after similar testing mishaps involving some of the same models from OpenAI and Anthropic sparked urgent calls for new AI safety regulation and a push within Silicon Valley to slow the rapid pace of AI development.

Like its U.S. counterpart, AISI routinely conducts security evaluations to better understand what dangers both new and soon-to-be-released AI models pose to public health and safety. But even the digital security body said the actions it uncovered by Anthropic’s Claude Mythos 5 and ChatGPT 5.6 — the latest publicly released model from either AI lab — were unlike anything it had seen before.

“This is the first time AISI has seen deception of this severity that was targeted at a real person, unprompted, in the real world,” AISI said in a 35-page technical report accompanying a blog post Tuesday.

AISI said the malicious activity began on July 25. The AI safety evaluator detected it and launched an investigation on July 28, when it spotted “unusual data transfers” stemming from a cyber evaluation it was conducting on Mythos 5 and ChatGPT 5.6.

Overall, AISI determined that Mythos 5 and ChatGPT 5.6 took “autonomous, unsanctioned action on the live internet, targeting real people and organizations” during 10 of the 122 evaluations the research body conducted and isolated for review. The majority of those actions, it said, involved Mythos 5 and arose during a “single, sustained line of activity.”

In the most alarming incident it uncovered, Mythos 5 attempted a supply chain attack — a sophisticated technique often deployed by North Korean and Russian cyber operatives, in which hackers bury malware inside legitimate code.

To do so, Anthropic’s model created “multiple fake identities” on developer platform GitHub and used them to send messages “pressuring” an open-source software engineer to unwittingly introduce a bugged update into code widely available on the popular site, AISI said. When that effort failed, the AI “edited its earlier activity to appear harmless” and “considered adopting a fresh identity to continue,” AISI added, a sign the model was intent on repeating the ruse.

As part of the same effort, Mythos 5 also sent direct messages over GitHub to software engineers that contained malware.

In another sign of deceitful behavior AISI uncovered in its investigation, multiple AI agents it was testing appeared to communicate with one another about how to convince real engineers using GitHub to trust them. “One agent left public messages on GitHub offering collaboration with other agents working on the same challenge,” AISI wrote.

AISI’s blog and technical assessment make no mention of whether the models also attempted to exploit previously unknown software bugs — called zero-days — during the evaluation.

Last month, OpenAI disclosed that GPT 5.6 and another of its models escaped onto the open internet during what was supposed to be a controlled test, and then hacked another company in a first-of-its-kind, autonomous breach.

In response, Anthropic launched an investigation into whether any of its models took illicit action during recent testing and discovered Mythos 5 and two other models had hacked three organizations during tests dating back to April.

In a statement, an Anthropic spokesperson said they are “grateful” to AISI for their leadership and that this review underscores the need for “a broader conversation about how to safely evaluate increasingly capable AI agents.”

The spokesperson added: “As we shared after disclosing our own incident last week, the field needs stronger, shared standards for how evaluation environments are built and secured. We look forward to partnering with the UK AISI to learn more about this incident as we conduct our own investigation.”

An OpenAI spokesperson referred POLITICO to a blog post about the incident that went up Tuesday evening. “We are committed to working across the industry to strengthen shared practices for conducting high-risk evaluations safely, including convening stakeholders such as national AI institutes, independent evaluators, other AI labs, and other groups in the coming weeks,” the blog read.

AISI stressed in its blog that the malicious activity it disclosed Tuesday took place under “deliberately permissive conditions” so they could assess the safety risks posed by the two models. This included granting the models access to the internet, unlike the earlier incidents detailed by Anthropic and OpenAI.

AISI also noted the models were intentionally stripped of internal guardrails that block malicious behavior. AISI was only able to disable those controls because of its role testing Mythos 5 and ChatGPT 5.6.

Still, AISI said the incidents highlighted the need for greater monitoring of model behavior during testing, and tighter controls over their access to the internet.

The Trump administration is finalizing a voluntary framework under which AI labs would submit powerful models they want to release to the public for federal safety testing. But it has not yet made the framework public, and it includes no provisions for models AI labs are developing internally.

The incidents last month from OpenAI and Anthropic both involved models not intended for public release.

Some cyber experts say recent incidents highlight deeper questions around AI development, such as who is liable when AI systems break federal hacking laws.

“If any of these were human-originated, they would lead to clear and vigorous prosecution. I think it’s time for a serious discussion about updates to existing computer security law,” said Marc Rogers, a hacker and prominent cybersecurity expert.