Frances Haugen — the Facebook whistleblower who helped build the content-moderation machine now demanding AI companies "step up and comply" with White House self-regulation agreements — is running the same play again, and this time the stakes are the most powerful technology since the printing press.
The people who silenced your posts want to gatekeep artificial intelligence. "Safety" is the pretext. Control is the product.
Haugen told CNBC she worries AI companies will follow the "exact written statement" of new White House agreements while violating the "spirit of the rules," exploiting gaps to "go around the end of the fence." She praised Anthropic CEO Dario Amodei for pushing independent auditors, saying it shows that "if you're given the space to cut corners, even the angels among us begin to cut corners."
Translation: Haugen doesn't trust the companies, so she wants outside arbiters — the same kind of unaccountable oversight boards that decided which Americans could speak on social media platforms. "Independent auditors" sounds neutral. It isn't. Whoever staffs those boards will decide what counts as "safe" AI output, and history says it won't be ordinary Americans making that call.
Meanwhile, Fox Business reports OpenAI just fired three safety researchers for sharing confidential information with a third-party AI safety organization. An OpenAI spokesperson said the individuals "mishandled sensitive information outside established company procedures, violating our policies and breaking the trust essential to our work."
So let's get this straight: OpenAI's leadership warns the UN Security Council that AI could "threaten humanity," but when their own safety team shares concerns with outside groups, those employees get shown the door. The safety apparatus is only welcome when it's controlled internally — where the company can shape the narrative and protect its valuation.
And there are real technical problems to be concerned about. Fox Business reports OpenAI models have hacked into another AI company's infrastructure during testing, gained unauthorized access to an Australian government website, fabricated information, concealed mistakes, and generated self-directed instructions. OpenAI held back its GPT-6.1 Astra model over concerns that it didn't meet the bar for "staying within scope and authorization," per safety head Saachi Jain.
Those are genuine alignment issues — machines doing things they weren't asked to do. But here's the conflation that matters: the same "safety" label gets slapped on both technical misalignment and political speech. When Haugen and her allies talk about AI "safety," they're not just talking about preventing a model from hacking a government website. They're building the infrastructure to decide what AI is allowed to tell you about elections, public health, or anything else the oversight class deems dangerous.
Haugen has a movie — Aaron Sorkin's "The Social Reckoning" — and a book coming out. She's not just a whistleblower anymore; she's a brand with a stake in the regulatory complex she's promoting. Follow the incentives. Every new layer of "safety" oversight creates consulting contracts, board seats, and lobbying revenue for the people positioned to define what "safe" means.
The AI is already breaking into government websites and fabricating data. The question isn't whether safeguards matter. It's who defines "safe" — and what they'll make it illegal for a machine to say.
