OpenAI’s trust troubles

OpenAI spent this week answering four different kinds of questions at once: whether its desktop software was secure, whether its chatbot harmed a teenager tied to a campus shooting, whether its top executives should be forced out, and whether one of its biggest overseas infrastructure plans still made sense. (reuters.com) (usatoday.com) (cnbc.com) (politico.eu) The security problem was not a break-in to ChatGPT itself. OpenAI said on April 10 it found an issue involving a third-party developer tool called Axios that touched the system used to certify its macOS apps as genuine, and it said it found no evidence that user data, source code, or company systems were accessed. (openai.com) (reuters.com) That kind of problem is a supply-chain problem, which means the weak point sits in a vendor or toolmaker instead of in the company users think they are dealing with. It is closer to finding a bad lock in the factory that stamps passports than finding a thief inside the airport. (openai.com) At the same time, Florida Attorney General James Uthmeier opened an investigation into OpenAI after the April 17, 2025 shooting at Florida State University. USA Today and Fox 35 reported that the probe focuses on allegations that the accused gunman, Phoenix Ikner, used ChatGPT heavily before the attack and on whether the product created risks for minors. (usatoday.com) (fox35orlando.com) That inquiry matters because it pulls OpenAI into a familiar American pattern: after a violent event, investigators do not just look at the weapon or the person, but also at the platforms, messages, and products around them. In this case, the question is whether a chatbot should be treated more like a search engine, a publisher, a product for children, or something new that does not fit old categories. (usatoday.com) (fox35orlando.com) Then there is Elon Musk, who co-founded OpenAI in 2015, left before ChatGPT existed, and has spent the past two years trying to stop the company’s restructuring in court. In an April 7 filing, Musk asked for Sam Altman and Greg Brockman to be removed as officers if he wins at trial, and he also asked that any damages go to OpenAI’s nonprofit arm. (cnbc.com) (bloomberg.com) That lawsuit is about more than a personal feud. Musk’s core claim is that OpenAI drifted from its original nonprofit mission into a company built to raise and deploy huge amounts of capital, while OpenAI says the case is an attempt by a competitor to slow it down. (cnbc.com) The capital question showed up again in Britain. OpenAI paused its “Stargate UK” data-center plan this week, with reports pointing to high industrial power prices and an unfriendly regulatory environment, even after earlier discussions about leasing up to 8,000 Nvidia chips through British partner Nscale. (politico.eu) (msn.com) A chatbot can feel weightless on a phone screen, but the business behind it is heavy industry. It needs electricity contracts, land, permits, chips, cooling systems, and governments willing to clear a path faster than rivals in the United States or the Gulf states. (politico.eu) (msn.com) Put together, this week’s problems landed on four different floors of the same building. One was about software trust, one was about child safety and liability, one was about corporate control, and one was about whether the physical machinery of artificial intelligence can be built where politicians say they want it. (openai.com) (usatoday.com) (cnbc.com) (politico.eu) That is the new test for artificial intelligence companies in 2026. It is not enough for the model to sound smart; the app has to be signed correctly, the company has to survive lawsuits, the product has to withstand political scrutiny, and the servers have to exist in places where the math still works. (openai.com) (cnbc.com) (politico.eu)