OpenAI and Anthropic Tell UK: Regulate Us

OpenAI and Anthropic told British lawmakers that they support regulation of advanced AI, a notable shift for an industry that has often warned against overly strict rules. The companies appeared at a sweeping parliamentary session alongside whistleblowers and security leaders, according to reports, and urged proactive oversight for frontier models. Their message was that governments should set guardrails before the most capable systems become widespread, rather than reacting after harm occurs. The hearing reflected growing pressure on policymakers to balance innovation with safety. Britain's spy chief warned that AI could be used to create new bioweapons, adding a national security dimension to the debate. Lawmakers are weighing how to regulate models that can be used for both beneficial and dangerous purposes, and how to do so without driving development offshore or stifling research. OpenAI and Anthropic's willingness to accept regulation may be partly strategic: clear rules can provide certainty, raise trust, and set standards that smaller competitors must also meet. But it also reflects genuine concern inside parts of the industry that safety efforts are lagging behind capability gains. The discussion included calls for testing, transparency, reporting requirements, and possibly licensing for the most powerful systems. Whistleblowers and security experts brought their own warnings about secrecy, rapid deployment, and the difficulty of auditing complex models. The hearing did not produce immediate legislation, but it signaled that the UK wants a role in shaping global AI governance. The key question is what effective regulation would look like. Too little could leave the public exposed to misuse, accidents, and disinformation. Too much, poorly designed, could entrench large companies and slow beneficial uses in medicine, science, and education. The companies' appearance suggests the debate has moved from whether to regulate frontier AI to how, when, and by whom. That is a significant change in the politics of AI, even if the practical answers remain unresolved.