TL;DR

The Commons Business, Innovation, Science and Trade Committee has opened a standing inquiry covering the adoption, governance and safety of AI, announced on 14 October, the day after its first evidence session with former OpenAI and Anthropic staff. Chair Liam Byrne said officials had admitted the government “has still not decided how or when it will act”. For UK firms, the practical thread to watch is how the committee tests the safety-testing and incident-reporting duties proposed in the Cyber Security and Resilience Bill.

The committee’s announcement frames the inquiry as a response to a week of public argument over AI safety, citing whistleblower testimony, resignations of safety specialists and warnings from national and international security figures.

The opening session

The inquiry began on 13 October with evidence from people who had left OpenAI and Anthropic over safety concerns. Serving representatives of Meta, Google DeepMind, OpenAI and Anthropic formed a second panel, which was asked to respond to those accounts and to describe its working relationship with AISI, the UK’s AI Security Institute. AISI itself appeared on the closing panel with senior government officials. We reported the pledges the labs made at that session on Wednesday.

The committee says it will measure the protections proposed under the bill against the evidence it gathers, including the companies’ internal safety-testing and incident-disclosure processes.

Byrne put it bluntly. “We cannot allow AI companies freedom to unleash dangers they cannot control,” he said, describing the first session as having exposed “disarray among the tech giants over the rules needed to keep us safe”. He tied the work to the UK’s G20 presidency and the chance to shape global governance standards.

An inquiry without an end date

Unusually, there is no one written-evidence call. The committee says the inquiry will stay open as individual strands conclude, so it can return to unresolved questions as events move. It plans to use evidence sessions on specific points, letters, visits and short reports, timed to land before ministers, regulators and industry make decisions, and to track whether its recommendations are acted on.

That structure matters for businesses with a stake in the outcome. Calls for evidence will arrive strand by strand, each with its own questions and deadlines, rather than in one window at the start.

Looking forward

In our view the design reads as a committee setting itself up as a permanent watcher while government policy on frontier AI remains unsettled, which is the gap Byrne’s own words describe. The first test will be the bill itself: if the committee concludes that the regime’s incident-reporting duties do not reach the behaviour the former lab staff described, that finding will arrive while ministers can still amend it. UK companies deploying frontier models should expect the inquiry’s questions about testing and incident disclosure to shape what their suppliers are eventually required to tell them.