UK AI regulation push amidst calls for global 'pacing' raises questions of impact

September 14, 2026, 18:24 GMT | Comment Calls for the UK to introduce dedicated AI regulation are gathering pace, but the latest proposals point in different directions on how risks need to be addressed.

Against a backdrop of influential tech giants calling for AI development to be slowed down, including Anthropic Chief Executive Dario Amodei’s headline-grabbing plan for industry-coordinated “pacing,” the different alternatives raise questions about who should be responsible for designing the guardrails and how much the UK can achieve by itself. 

The Joint Committee on Human Rights, a cross-party group of UK lawmakers, added to the pressure built over the weekend with its own report published on Monday morning. It calls for an AI bill, the creation of an independent oversight body and stronger powers for the influential UK AI Security Institute, or AISI (see here). Its recommendations go further than the government’s current reliance on sectoral regulators, which the committee says offer fragmented protections.

The model recommended by the group is also broader than what was proposed last week by Labour lawmaker Alex Sobel, with the support of campaign group ControlAI, in a bill that seeks to prohibit the development of superintelligent AI in the UK (see here).

Michael Birtwistle, associate director at the Ada Lovelace Institute, said the committee’s report and the calls for a ban share a concern that companies should not be left to decide which risks others must accept.

But he warned that a bill banning superintelligence would not, by itself, stop the risks that are already arising from the currently available AI models: agents launching cyberattacks; chatbots influencing political opinion; and harms to children or discriminatory recruitment systems, for example.

The UK government has so far resisted introducing AI-specific legislation, even if Labour's 2024 general election manifesto said it would regulate AI risks.

But, in a new statement to MLex, a government spokesperson said ministers should not assume that the existing framework would “always be sufficient as AI capabilities develop.”

“We take the risks from advanced AI extremely seriously and recognize the importance of ensuring our approach keeps pace with a rapidly evolving technology,” the spokesperson said, adding that any future measures would be "evidence-based and focused on the risks that need to be addressed."

The government also said that, because the risks cross borders, it is working with the companies developing the most advanced systems, allies and international partners. That includes using the AISI to build a scientific understanding of advanced models and their risks, it said.

This government comment follows AI minister Kanishka Narayan’s written statement last week which said the government would consider whether protections for increasingly autonomous systems should be clarified or strengthened through the Cyber Assessment Framework, a forthcoming statutory code of practice or National Cyber Security Centre guidance.

It also follows the government’s rejection last week of Sobel’s private members’ bill, when a spokesperson said ministers were exploring targeted interventions to address the most significant AI-related national-security risks instead of the bill's proposed ban on superintelligent AI (see here).

The UK director of campaign group PauseAI, Joseph Miller, said that while it's good that the British government signaled the possibility of targeted interventions, this also suggests ministers are waiting for some risks to materialize.

"You will never get definitive proof that AI can cause human extinction until it's too late. You need to look forward, listen to the experts, anticipate the progress, notice the trend, and then get ahead of the situation," Miller told MLex.

"Currently the government has shown that they are willing to act on the risks that have already been clearly demonstrated, but they currently don't seem to be at all capable of getting ahead of the risks that are coming down the line," he said.

— Pacing AI — 

The UK’s arguments are now unfolding alongside a much wider debate on whether AI development itself should be slowed down by regulatory frameworks and, if so, how this could be achieved globally.

In an essay published Saturday, Anthropic's Amodei called for “pacing” the frontier so that safety work can keep up with gains in AI capability. He said the most effective route to achieve this would be "via regulation that targets all US frontier AI companies," focused on transparency and third-party auditing, alongside permanent embedded evaluators equipped with employee-like access to the leading AI labs.

Amodei also argued that while legislation is being developed across the world, frontier companies should coordinate voluntary standards and limits between themselves, with an antitrust waiver to avoid collusion accusations.

"For antitrust reasons, it’s helpful for the US government to mediate or at least enable these discussions — they don’t need to participate, but do need to issue a narrow waiver for certain kinds of safety conversations," Amodei wrote, prompting the EU to say that safe AI rollout isn't hindered by its antitrust rules (see here).

Amodei's proposal falls short of the full pause sought by some campaigners, although the Anthropic boss said such efforts are worth "floating," even if they're "unlikely to actually happen any time soon."

Later on Monday, Anthropic rival OpenAI said it would support stronger UK rules for the companies developing the most powerful models (see here), as long as they don't add burdens for startups and smaller developers.

"We’re ready to work with the UK government and other partners to help people benefit from AI safely," said the company's head of EMEA policy, Tom Duff Gordon.

Andrew Strait, an AI researcher who recently left a leadership role with the AISI, said there were reasons to be skeptical when the largest AI companies in the world proposed coordinating development and writing the standards against which they would be judged.

Similar to Ada Lovelace's Birtwistle, Strait warned against narrowing the concept of AI safety to alignment, biological threats and cyberattacks, leaving aside child safety, psychological harms, discrimination and economic impacts.

United Nations High Commissioner for Human Rights Volker Türk made a similar point in an open letter Monday. Industry co-regulation is “an urgently needed first step,” he said, “but on its own, it is nowhere near sufficient.”

Countries hosting frontier companies should require independent verification and human rights due diligence, the senior UN official argued.  "They need to align their approaches so that competition between jurisdictions does not encourage a race to the bottom."

The UK’s renewed legislative pressure could shape that global discussion, PauseAI's Joseph Miller said.

"We definitely need an international treaty and the best way the UK can help achieve that is to lead by example. If the UK makes a clear stance that we need a pause on AI development, then that raises the question for the US," he argued.

But there might be practical limits to UK action. Keegan McBride, director of science and technology policy at the Tony Blair Institute, told MLex last week that the AISI is a “fantastic institution,” but cannot provide a strategic advantage over the US if the UK is not building or using frontier systems itself (see here).

Miller said that while the UK is not leading the development of frontier AI, it has "a relative strength in AI safety."

"London is probably the biggest hub of AI safety outside of San Francisco," Miller said, echoing other experts and mentioning the role of the UK AISI.

"I'm fairly optimistic that the UK government will continue to invest in the AISI. But if they don't support it with legislation that gives them mandated access to models, then it will fade into irrelevance," he warned.

With Trump pushing back against the calls for AI guardrails (see here) and the prospect of a meeting convened by King Charles with some of world's leading AI labs later this week, the question of how the UK will position itself if it does move ahead with wide-ranging AI regulation is becoming increasingly thorny.

Please email editors@mlex.com to contact the editorial staff regarding this story, or to submit the names of lawyers and advisers.