LLMs

OpenAI's Safety Culture Is Fracturing in Public

Three fired OpenAI safety researchers disputed misconduct allegations this week in an open letter, warning that their dismissals are sending a chilling message to anyone inside the company who raises concerns about AI risk. Days later, a separate safety employee resigned, calling the company's culture 'broken.' The back-to-back departures arrive as OpenAI's PR team was caught on tape trying to redirect a Vanity Fair editor away from questions about a ChatGPT user's suicide, a moment that crystallized how aggressively the company is managing its public image on safety topics. Taken together, these incidents point to a deepening tension between OpenAI's commercial velocity and its original safety mission. The company simultaneously published a trust and transparency report on government data requests and child safety, which reads less as reassurance and more as a sign that scrutiny is intensifying. Anthropic, meanwhile, updated its usage policy this week to explicitly ban model abuse and election interference, a quiet but pointed contrast. For enterprise buyers and regulators watching this space, the question is no longer whether AI safety culture is under strain at the frontier labs; it's whether anything structural will force a change.