OpenAI’s Trust Problem Isn’t Just About AI Safety—It’s About Sam Altman

OpenAI has a split personality problem.

On Monday, the company published a slick policy document about how it plans to keep humanity safe when superintelligence arrives. It talks about staying “clear-eyed” about risks, monitoring for AI systems that evade human control, and preventing governments from using AI to undermine democracy. The usual corporate reassurance boilerplate, but dressed up in earnest language about “higher quality of life for all.”

On the exact same day, The New Yorker dropped a massive investigation into whether Sam Altman can be trusted to follow through on any of it.

Reading them side by side is almost surreal. One document paints a picture of a responsible steward of powerful technology. The other paints a picture of a CEO who, according to people who’ve worked closely with him, has a credibility problem that runs deep.

The New Yorker piece isn’t subtle. Multiple OpenAI insiders reportedly expressed deep reservations about Altman’s trustworthiness. The headline basically says it all: “The problem is Sam Altman.” Not the technology, not the competition, not the regulatory landscape. Him.

This matters because OpenAI’s entire pitch to the public and to regulators rests on trust. The company has positioned itself as the responsible AI lab, the one that will put safety first even if it means slowing down. But if the people inside the building don’t believe their own CEO can be trusted, why should anyone else?

I’ve been around long enough to know that corporate policy papers are written by comms teams, not by the CEO’s inner circle. But the timing here is brutal. You can’t release a document asking the world to trust you with superintelligence on the same day your own employees are telling journalists they don’t trust you.

The disconnect between OpenAI’s public messaging and its internal reality has been growing for a while. The boardroom drama last year, the rapid product releases that sometimes felt rushed, the constant pivots on safety commitments. Each incident chips away at credibility. This New Yorker piece feels like a crowbar.

OpenAI’s policy paper acknowledges that without proper mitigation, “people will be harmed.” It’s the kind of statement that sounds responsible until you realize the person in charge of those mitigations is the same person people inside the company don’t trust to follow through.

The real question isn’t whether OpenAI can build superintelligence. It probably can. The real question is whether anyone will believe the company when it says it’s handling that power responsibly.

Right now, the answer looks like no.