Whether regulation is keeping pace with the rapid development of AI is increasingly at the centre of public and regulatory debate. Since the summer, we have seen major security incidents, dire warnings from leading AI executives, and renewed calls for global cooperation on AI governance.
Warnings from the AI industry
Soon after a senior safety researcher at Anthropic put the chance of AI ending humanity within the next decade at more than 10%, Dario Amodei, the CEO of Anthropic, published an essay urging the AI industry to slow the pace at which AI model capabilities improve. He also warned of the serious harm that could be caused by misaligned AI agents, as illustrated by the recent Hugging Face incident. In that incident, OpenAI’s AI agents gained unauthorised access to Hugging Face’s systems during a cybersecurity test, and the non-profit Legal Advocates for Safe Science and Technology (LASST) has since sued OpenAI.
Among the measures proposed by Amodei was coordination among leading AI companies on common safety standards and the pace of AI development, with governments mediating or enabling those discussions, together with independent evaluators verifying compliance with safety commitments. The proposal by Anthropic’s CEO also included a controversial request: a narrow antitrust waiver so that AI competitors are allowed to coordinate on safety measures. Demis Hassabis (from Google DeepMind), Elon Musk (from xAI) and Sam Altman (from OpenAI) showed public support for Amodei’s message. Both Hassabis and Altman had previously advocated centralised standards and testing by independent third parties, amongst other measures.
Separately, OpenAI has been openly critical of the UK’s current sector-specific approach to AI regulation. The company has argued that the UK’s reliance on sectoral regulations is no longer sufficient and has called for binding rules for the handful of companies developing the most powerful AI systems, with the AI Security Institute (AISI) playing a central role. This is significant given ongoing concerns regarding the AISI’s current role, particularly the fact that model testing remains voluntary. The limitations of that approach were brought into sharp focus when Anthropic did not offer the AISI full access to Claude Mythos 5.1 for testing prior to its release, which Anthropic told MPs on 13 October 2026 was a matter for the US government, adding that the AISI now has access. At the same session, representatives from OpenAI and Anthropic told MPs that reliance on voluntary safety commitments is no longer sufficient for the most powerful AI systems.
Government responses
UK
Pressure to reform the UK’s AI regulatory framework is growing. A report on Human Rights and The Regulation Of AI was recently published by the Joint Committee on Human Rights (JCHR). The report was very critical of the existing regulatory framework. It called it “patchy and confused” and advocated for the introduction of new comprehensive AI-specific regulation, going further than the more targeted rules proposed by OpenAI. The report calls for new legislation governing AI systems from development through to deployment and use. It also recommends the establishment of an independent AI supervision body, enhanced safeguarding measures for automated decision-making, and placing the AISI on a statutory footing with stronger powers.
The government has not yet published a consultation on an AI Bill, despite Labour’s 2024 manifesto commitment to binding regulation for developers of the most powerful AI models. Should the government introduce dedicated AI legislation, it is likely to extend beyond questions of frontier-model safety and encompass wider issues relating to the deployment and use of AI systems.
EU
In her annual State of the Union address to the European Parliament, European Commission President Ursula von der Leyen confirmed plans to engage with leading frontier AI developers on how the EU can support the industry in slowing down the pace of AI development, as called for by major AI executives. Von der Leyen also expressed her intention of collaborating with partner countries (such as the UK and Canada) on AI security and assessment of AI models.
US
President Trump has repeatedly stated his focus on beating China in the global race for AI leadership, describing warnings that AI could take over the world or destroy humanity as a “hoax”. Meanwhile, Andrew Ferguson, Chairman of the US Federal Trade Commission (FTC), expressed strong scepticism about AI companies pursuing antitrust exemptions while simultaneously advocating for additional regulation, arguing that such an approach could reinforce incumbents’ market positions by raising barriers to entry for smaller competitors. Moreover, despite ongoing legislative activity at state level, federal AI regulation remains improbable. This is particularly so in light of the voluntary framework for pre-release government access to “covered frontier models” established by President Trump’s June 2026 executive order.
What does this mean going forward?
For the moment, the EU AI Act is the most comprehensive piece of AI legislation in the world, and cyber agencies and sector regulators across the UK and EU are increasingly focused on the risks posed by AI. New AI laws may be introduced, especially in the UK, but immediate compliance concerns continue to dominate the agenda for many organisations.
For UK organisations, the immediate challenge is navigating an increasingly uncertain and fragmented regulatory environment. Businesses operating across the UK and EU are already preparing for compliance with the EU AI Act, while also monitoring the possibility of future UK AI-specific legislation. Organisations developing, deploying or procuring AI systems may therefore need to invest in governance, risk management, testing and documentation processes that can accommodate differing regulatory expectations across jurisdictions. Even in the absence of new UK legislation, growing scrutiny from regulators, customers and business partners means that robust AI governance is becoming a commercial necessity as well as a compliance requirement.
As concerns mount regarding the increasing capabilities of AI agents and their potential for harm, support is growing for stronger oversight of frontier AI models, especially among key industry figures. Calls for international standards and cooperation that keep up with the rate of AI technology development, and for evaluation of AI models by independent third parties, are recurring themes in the emerging debate. Whether such measures can be implemented effectively remains uncertain, particularly as governments and technology companies face growing incentives to prioritise innovation, competitiveness and strategic advantage.
Please contact Jose Saras and Xavier Prida for detailed advice on the above.
The material contained in this article is only for general review of the topics covered and does not constitute any legal advice. No legal or business decision should be based on its content.
This article is written in the English language. Preiskel & Co LLP is not responsible for any translation of all or part of its content into any language.