OpenAI hit by security and safety incidents
OpenAI disclosed a security issue tied to a third‑party macOS certification tool and separately paused its Stargate UK data‑centre project; authorities also arrested a suspect after an incendiary attack linked to threats at the company. Together these developments have shifted mainstream AI coverage toward security, infrastructure strain and political risk. (reuters.com) (independent.co.uk) (cnbc.com)
OpenAI spent Friday answering three very different problems at once: a software supply-chain scare on Apple computers, a stalled British data-center buildout, and a criminal case after a firebomb was thrown at Sam Altman’s San Francisco home. (cnbc.com) (independent.co.uk) The software issue started with a third-party developer tool called Axios that helps certify OpenAI’s macOS apps as real OpenAI software and not fakes. OpenAI said it found no evidence that user data, company systems, or intellectual property were accessed. (usnews.com) That kind of certification works like a tamper seal on a medicine bottle: your Mac checks the signature before it trusts the app. OpenAI said it is rotating certificates and told users to update ChatGPT, Codex, Atlas, and Codex Command Line Interface on macOS. (ithinkdiff.com) The second problem is physical infrastructure, which is less visible but just as important for artificial intelligence. Training and running large models requires giant data centers packed with advanced Nvidia chips and huge amounts of electricity. (cnbc.com) (politico.eu) OpenAI had planned a British version of its Stargate project with Nvidia and London-based Nscale, including sites at Cobalt Park near Newcastle and at Blyth in northeast England. The plan included exploring leases for up to 8,000 advanced Nvidia chips in early 2026. (independent.co.uk) (politico.eu) This week OpenAI put that project on hold, citing high energy costs and an unfavorable regulatory environment in the United Kingdom. That pause landed awkwardly for Prime Minister Keir Starmer’s push to sell Britain as an artificial-intelligence hub through designated “AI Growth Zones.” (independent.co.uk) (cnbc.com) Then came the security threat in San Francisco. On April 10, police arrested a suspect after a Molotov cocktail was allegedly thrown at Altman’s house and after threats were made outside OpenAI’s headquarters. (cnbc.com) (nbcnews.com) San Francisco police said officers responded around 3:43 a.m., and OpenAI said no one was hurt. The case turned a company already used to online backlash into a target of offline violence. (cnbc.com) (nbcnews.com) Put together, the week showed how an artificial-intelligence company can be squeezed from three sides at once: software trust on users’ laptops, power and permits for the buildings behind the models, and personal security for the people running the company. Those are now part of OpenAI’s day-to-day business in the same way model launches used to dominate the story. (independent.co.uk) (cnbc.com)
