OpenAI backs down on aggressive exit agreements

Silicon Valley equity has an unwritten rule: you take below-market pay, put in crazy hours, and when you leave, your vested equity is yours to keep.

OpenAI apparently had a very different set of rules, until public outrage forced a massive, humiliating retreat.

The clawback trap and the paper trail

On Thursday, OpenAI sent an internal memo to former and current employees confirming it will no longer enforce aggressive non-disparagement and non-solicitation agreements, CNBC reported.

Under the controversial exit documents first broken by Vox, departing staff were given an impossible choice: sign an NDA with a lifetime non-disparagement clause, or forfeit all of their vested equity, which in some cases was worth millions of dollars.

In the internal memo, OpenAI promised it has not and will never cancel vested units, regardless of whether former staffers signed the release, and that standard non-disparagement terms will be removed entirely from departure paperwork.

When the story first broke, CEO Sam Altman took to X to play dumb, claiming he was genuinely embarrassed and had zero knowledge that the clawback clause even existed.

That defense collapsed fast. Follow-up reporting from Vox revealed leaked incorporation documents showing both Altman and Chief Strategy Officer Jason Kwon had personally signed off on language giving the startup nearly arbitrary authority to seize equity or block sales.

Outgoing employees reported being pressured by HR with strict seven-day deadlines, warned that refusing to sign would directly hit their equity payout.

Why it matters

For current and former OpenAI staffers, this memo is an overdue sigh of relief. Holding someone’s hard-earned equity hostage to buy lifetime silence is legally questionable and ethically rotten. Ex-employees can now speak up without fear of financial ruin.

For the AI ecosystem, we think this is another massive hit to OpenAI's public credibility. The retreat comes during one of the rockiest weeks in company history, right after pulling the Sky voice following Scarlett Johansson’s public outrage and completely disbanding its Superalignment team after leaders Ilya Sutskever and Jan Leike walked out.

Altman’s personal brand is arguably the biggest casualty here. When the board fired him last year, their official complaint was that he wasn't consistently candid. Claiming on social media that you had no idea about aggressive clawbacks, only for leaked documents to show your signature right on the page, feeds straight back into that exact narrative.

In the war for elite research talent, OpenAI was already losing safety-minded minds to rivals like Anthropic. Showing the world that you're willing to weaponize compensation to bury internal dissent makes recruiting top-tier researchers that much harder.