The Newshound

MI5 names AI misuse a national security threat as tech giants face MPs over bioweapon fears

Daniel Kokotajlo, who told MPs that employees at OpenAI were pressured to minimise risks and that internal safety research was suppressed before publication.
Daniel Kokotajlo, who told MPs that employees at OpenAI were pressured to minimise risks and that internal safety research was suppressed before publication.

Senior executives from OpenAI, Anthropic, Google DeepMind and Meta faced tough questioning from Parliament's business, innovation, science and trade select committee on Wednesday over AI safety and governance. Anthropic warned MPs that AI could help bad actors "create biological or chemical weapons," while MI5 named AI misuse as a direct national security threat. OpenAI revealed it had scrapped the release of its latest model, GPT-6.1 Astra, over safety concerns, saying it "didn't refuse enough." The hearing exposed the raw tensions between the race to deploy AI and the risks it poses to public safety.

OpenAI's Tom Duff Gordon told MPs the shelved GPT-6.1 Astra model 'hadn't progressed enough with alignment.'

Anthropic's Lord Booth-Smith warned the firm is finding tens of thousands of cyber vulnerabilities a month through its Mythos system.

MoreLBCOpenAI ‘not racing to the bottom’ on AI safety, MPs toldCNAAI giants face UK parliament over safety and regulation - CNAYorkshire Post'We are not racing to the bottom': OpenAI takes safety really seriously, MPs toldLondon South EastOpenAI 'not racing to the bottom' on AI safety, UK MPs told
Read today's full feed at The Newshound →