Former AI researchers have warned MPs that the development of advanced AI is moving faster than efforts to make it safe. Their testimony raises a pointed question for Parliament: who sets the limits when the companies building these systems are also judging the risks?
Watch Desk analysis
What happened
CityAM reports that Daniel Kokotajlo, a former OpenAI researcher, told the Business, Innovation, Science and Trade Committee that he had lost confidence in OpenAI’s willingness to act responsibly as it pursued superintelligence. He described the industry’s risk tolerance as “catastrophically high”.
The report also quotes former OpenAI researcher Steven Adler saying there had been resistance within the company to describing its technology as dangerous. Adler acknowledged that the company’s internal culture may have changed since he left in late 2024. Geoffrey Irving, a former researcher at OpenAI, Anthropic and the UK AI Security Institute, told MPs that capability advances were not being matched by safety research. CityAM says OpenAI and Anthropic were contacted for comment.
Why it matters
These are witnesses’ assessments, not proof that a particular system is unsafe or that the companies acted improperly. But testimony in a parliamentary inquiry gives lawmakers a chance to test what researchers say they saw against the companies’ account and the evidence behind their safety practices.
The distinction between a warning and a demonstrated failure matters. So does the practical question the witnesses put before MPs: whether competition between developers can push acceptable risk higher unless governments establish a clearer ceiling.
Our read
The most useful part of this hearing is not the most alarming phrase. It is the disagreement now on the record about whether safety work is keeping pace with capability, and who should be responsible for making that judgement. Parliament should press for specifics: what risks, what safeguards, and what evidence would change the witnesses’ view.
What to watch
- Whether OpenAI and Anthropic respond to the specific claims raised in the hearing.
- What evidence MPs request about safety research, internal decision-making and risk thresholds.
- Whether the committee recommends limits or other measures for advanced AI development.
Discussion spark: Should governments set binding risk limits for advanced AI developers, or should companies retain the lead role in deciding when a system is safe enough to release?
Sources and evidence
- Source update (13 October 2026, 14:49 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.