In Westminster, OpenAI and Anthropic are calling on the UK to introduce binding regulations for the most powerful models

Speaking before British MPs, the two laboratories advocated for mandatory pre-market testing and a liability regime, tailored to the capabilities of the models.

On October 13, 2026, OpenAI and Anthropic spoke in the House of Commons in favor of binding regulation of AI in the United Kingdom, alongside Google DeepMind and Meta. The two laboratories are advocating for obligations proportionate to the capabilities of the systems rather than uniform rules: mandatory assessments of dangerous capabilities before any powerful model is put on the market, and a framework for liability in the event of a problem.

So far, London has relied on voluntary tests conducted by the AI Security Institute, while Labour promised in its 2024 programme to legislate on advanced models. The session also heard from MI5 Director-General Sir Ken McCallum, who spoke about the risk of AI facilitating the development of biological weapons and announced the creation of an AI threat assessment centre.

Capacity-based thresholds would mainly affect large laboratories, which have the means to comply with them; there is still the fear that an AISI that has become a mandatory passage could turn into a bottleneck.