1. Home
2. News
3. OpenAI

Fired OpenAI safety researchers say colleagues are now afraid to speak up

Their letter asks OpenAI to keep its models readable and to hold Altman to his promise of outside auditors.

By Think Facility · October 8, 2026 · OpenAI AI safety Save

On October 8, 2026, the three safety researchers OpenAI fired the week before published an open letter to the company's safety oversight bodies. Tomek Korbak, Jasmine Wang and Mikita Balesni say they acted within OpenAI's working norms, and that the way their firing was handled has left former colleagues afraid to speak.

Wang posted on X that afternoon (the first post of a thread). She gives the one reason she says OpenAI gave her.

What the letter says about the firing

The letter goes to OpenAI's Safety and Security Committee, its Safety Advisory Group and its Mission Advisory Council. The three write that they weren't the source of a leak to The Information, and that they didn't deal with outside groups beyond what their jobs called for.

Their bigger worry is the precedent. If conduct that counted as normal last month can now get someone fired on the spot, they write, nobody at OpenAI knows where the line is. They describe the freedom to work closely with outside safety experts, without fear, as a safety mechanism in its own right.

All three have been in the field for years. Balesni was a founding member of Apollo Research, Korbak worked at Anthropic before OpenAI, and Wang led a team at UK AISI (the letter uses the acronym) before returning to OpenAI in 2025. The letter also credits Wang with coining "pacing", the word behind the Pacing the Frontier petition it says 394 OpenAI employees signed.

Three parting requests

The first is that OpenAI keep embedding outside safety auditors and keep working with groups like METR. The letter cites what Sam Altman posted on September 12.

The second is about chain-of-thought monitorability. Two of the three were lead authors on a cross-industry paper on it before they joined OpenAI. Korbak, the letter says, worked on finding out why monitorability dropped in OpenAI's Astra-class models.

They want OpenAI to keep its frontier models monitorable, and to back industry coordination so labs don't race into designs nobody can read. That's a request to the whole field, really, and the letter frames it that way.

The third asks OpenAI to say publicly that staff can still raise concerns inside and outside the company, and to spell out how they may work with outside safety groups.

OpenAI's side

We couldn't find a statement on the firings on OpenAI's own site or accounts. Its position has reached us only through press reports, so it isn't quoted here.

The letter is signed Tomek, Jasmine, and Mikita. It closes by urging the colleagues they left behind to keep holding OpenAI to its mission.