An OpenAI Engineer Charged With Bugging the Room, Two Labs Ask Britain for Binding Rules, the Industry War-Games the Day After

An OpenAI engineer is charged with hiding a recorder in an executive conference room. Two days later the two biggest AI labs ask the British Parliament to bind them by law, while their two biggest rivals say no. And the same companies are privately rehearsing the day the public turns on them.

On Tuesday, October 13, the US Department of Justice announced that a Saudi national employed at OpenAI had been arrested and charged with planting a recording device in a company executive conference room. On or around September 30, the company found a sophisticated recording device inserted into the back of a television in an executive conference room in one of its San Francisco offices. OpenAI determined the device had been installed about two days earlier, concealed in toilet paper, and its contents were encrypted. The employee, Abdulrahman Alfozan, 35, had been arrested on Friday, October 9, and appeared in federal court on Tuesday. OpenAI said it had used its own models to work out where the device came from and identify a specific employee, and that it had found no indications the man inappropriately accessed or exfiltrated model weights, source code or customer data. 1

That same Tuesday, four frontier AI companies gave evidence to the House of Commons Business, Innovation, Science and Trade Committee. Anthropic's head of Europe, Liam Booth-Smith, told MPs the rules should be binding and that voluntary commitments were not enough. OpenAI vice president Tom Duff Gordon backed a statutory framework and an international convention, though he said mandatory requirements should begin with incident reporting. Their rivals would not follow: asked directly whether he supported a statutory framework, Google DeepMind's Owen Larter said the company would focus on standards, and Meta's Rob Sherman pointed to Britain's existing regulators, telling MPs they were not writing on a blank page. The same session heard how little the current system compels. Britain's AI Security Institute tests frontier models before release only when companies choose to hand them over, and its director, Henry de Zoete, acknowledged he had no way to stop a company releasing a model the institute judged too dangerous. Asked whether the whole regime was voluntary, the Cabinet Office's top civil servant, Cat Little, said: "At the moment, yes, it is." A few hundred metres away, MI5 director general Ken McCallum used his annual threat update to warn that AI could change the stakes on existing or new bioweapons. Committee chair Liam Byrne said Britain should use its coming G20 presidency to broker an international convention: "The risks are multiplying faster than our safety regime can control." 2

Privately, the same companies are getting ready for worse. Axios reported on October 9 that top executives at Anthropic, OpenAI and other AI companies are gaming out scenarios for a public and political revolt after a catastrophic AI event, most likely a cyberattack that shuts down access to financial services, internet connectivity, or even power and water. An OpenAI spokesperson said the company runs preparedness exercises and does not treat the scenarios as inevitable; Anthropic declined to comment. Many AI industry insiders told Axios they believe a major event will occur within the next six to 12 months, and the companies are racing to educate members of Congress on the legislation and policies Washington will reach for after a first catastrophic event. 3