OpenAI ‘not racing to the bottom’ on AI safety, MPs told

OpenAI has said it is not “racing to the bottom” on AI safety as MPs questioned bosses from technology firms over the safety and security of their AI agents.

Meanwhile, Claude parent firm Anthropic suggested that there are risks that the fast-changing technology could be used by bad actors to “create biological or chemical weapons”.

It came as the Parliament’s business, innovation, science and trade select committee questioned executives from OpenAI, Anthropic, Google DeepMind and Meta.

Tom Duff Gordon, vice-president and head of Europe, Middle East and Africa policy at OpenAI, said the group scrapped the release of its latest model, GPT-6.1 Astra, over safety concerns.

He stressed the decision to halt the model’s release highlighted a desire to maintain safety as it develops its technology.

He said: “We found it didn’t refuse enough, it hadn’t progressed enough with alignment.

“It is evidence that OpenAI takes safety really seriously, that we are not racing to the bottom.”

Lord Booth-Smith, head of Europe, Middle East and Africa policy at Anthropic, stressed that the company has “rigorous” safety processes.

He said: “For our deployed models, we think they are safe and we have mitigations in place.

“But it is also worth pointing out that we are learning a lot as we go about how these models are used.

“In Anthropic right now the surface area of immediate risk is definitely in the cyber domain. Mythos has been finding tens of thousands of vulnerabilities a month.

“It also makes sense that this surface area covers biology and other areas.

“It is possible that real harm could be caused in the space of helping people to create biological or chemical weapons.”