OpenAI Fired Three Safety Researchers. Their Open Letter Says the Lines Are Shifting.

Three firings, two very different stories

Weeks after pledging to open its doors to outside safety experts, OpenAI fired three of its own safety researchers, saying they mishandled sensitive information. The Rundown AI reported on October 9 that all three have now gone public with an open letter, denying the claims and warning that the firings are chilling OpenAI's safety culture.

The timing is what makes the story sting. OpenAI had just promised more outside scrutiny of its safety work. Firing three of the people who do that work, weeks later, was always going to be read as a message, whatever the internal facts turn out to be.

The researchers, Tomek Korbak, Jasmine Wang, and Mikita Balesni, say they acted within their job mandates. Their sharpest point is about retroactive rules: if conduct that used to be normal can become grounds for firing, staff are left guessing where the line sits on any given day.

What each side is claiming

The letter's title makes its ask plainly: OpenAI cannot make AI safe on its own. The researchers argue that outside eyes are a feature of the safety system, not a threat to it. The claims, side by side:

• The trio denied leaking a story about less monitorable AI and said their outside work stayed within policy, with leadership informed
• Wang says her access to an executive's email was delegated for recruiting, and that she reported opening a sensitive message within minutes
• The letter urges OpenAI to protect its open culture and keep outside auditors embedded in safety work
• OpenAI told TechCrunch the firings were not about raising concerns, and pointed to a pattern of misconduct

Why builders should care about someone else's HR fight

Most of us will never work at OpenAI, but all of us build on platforms whose rules move. Anyone who has had a video flagged under a policy that read differently last month knows the feeling the letter describes. Shifting lines do not just punish the people caught by them. They teach everyone watching to stop taking the small risks that good work requires.

There is history pressing on this too. The Rundown connects the letter to the fallout from OpenAI's disbanded Superalignment team and a recent safety exit over a culture the departing researcher called broken. Each episode on its own is a personnel story. Together, they form the pattern the letter is really about.

The practical lesson travels well: write your standards down, apply them the same way on good days and bad ones, and when the line has to move, say so before someone crosses the old one. Culture is just the set of rules people believe will still be true next month.