Three OpenAI safety researchers say their firing signals a broader shift that could stifle safety work and collaboration with external experts.
Three OpenAI safety researchers say they were fired last week for sharing information with a third-party safety group, and they’ve pushed back with an open letter that accuses the company of creating a chilling effect. OpenAI says the firings were due to policy violations around handling sensitive information. The researchers warn the firings could scare others away from safety discussions, potentially slowing progress toward safe AI. OpenAI stresses its commitment to safety and external oversight, while details about the policy violations remain unclear. This dispute highlights a broader debate over how much internal versus external review AI labs should have, and what protections whistleblowers need as AI systems grow more capable. Look for updates on how OpenAI will address these concerns in the coming weeks.
Three OpenAI safety researchers say they were fired last week for sharing information with a third-party safety organization, and they’ve pushed back with an open letter that accuses the company of creating a chilling effect. The researchers, Jasmine Wang, Tomek Korbak, and Mikita Balesni, say ongoing collaboration with outside experts has long been a key part of handling risky AI work, and they warn abrupt terminations could scare others from engaging in safety conversations.
OpenAI has not laid out the exact policy violations behind the firings, but the company has previously said the researchers violated rules about accessing and handling sensitive information. In response to the letter, OpenAI circulated an internal memo attributed to a research leader praising the researchers while stopping short of saying the firings were retaliation. The memo reiterates the company’s commitment to safety work and external oversight.
In their open letter, the researchers argue that advanced AI safety requires ongoing research and outside input. They describe OpenAI as “not a normal company,” and warn that the current atmosphere could deter employees from raising safety concerns or working with external evaluators. They say the culture of open dialogue has been a safety mechanism for AI work in the past.
In a thread on X, Wang provided more detail about one firing reason. She says she was told she’d been dismissed for accessing an executive’s email. She explains she asked IT to remove access after she no longer needed it, but the inbox remained accessible and a sensitive email was opened by mistake. She adds that the reasons cited by the company do not add up and says she acted in good faith within the norms in place at the time.
The letter urges OpenAI to keep embedding third-party safety auditors and to preserve dialogue between safety researchers and the broader safety ecosystem. The researchers claim that punishing employees for raising concerns or collaborating with outside evaluators could hinder progress toward safe AI deployment.
What this means for OpenAI’s safety program and for the wider AI field is still unfolding. OpenAI says it values safety work and external oversight, while details about the policy violations remain unclear. Both sides acknowledge ongoing questions about how internal and external review should balance in frontier model development.
OpenAI fired me last week, along with two of my safety colleagues. I was given one reason: that I accessed an executive's email. I want to say this plainly, because too many of OpenAI’s history is smoke and mirrors when people disappear: https://t.co/RC4fm2nV95
— Jasmine Wang (@j_asminewang) October 8, 2026
The researchers say they were dismissed after participating in safety work with outside groups and sharing information they believed was necessary for assessments. They emphasize this was not about avoiding scrutiny, but about maintaining a process that lets safety researchers collaborate with external evaluators without fear of retaliation.
OpenAI has described the firings as policy violations related to handling sensitive information, but has not publicly detailed exactly what policies were breached or whether that was the sole cause.
For readers, the big question is how AI labs safeguard safety, privacy, and fairness when researchers push for risk mitigation. The open letter argues that a culture of fear around safety work can push important discussions underground. OpenAI says it remains committed to safety and third-party oversight, but the right balance between internal checks and external review is under debate.
The writers say abrupt terminations erode the value of open safety work and external accountability. OpenAI counters that it does not fire employees for raising concerns and that safety remains a priority alongside external audit processes.
Outside experts say independent oversight helps catch issues early. If firings dampen collaboration with safety evaluators, some fear it could slow efforts to keep advanced AI systems safe and trustworthy.
Right now, both sides offer limited detail about the policy issues behind the dismissals. The open letter calls for renewed emphasis on safety collaboration, while OpenAI emphasizes its ongoing safety program and external auditing. The story next moves to how OpenAI responds and whether new protections or procedures are announced.
What sparked the dispute?
The researchers say they were fired after sharing information with a third-party safety group; OpenAI says the firings were for policy violations around handling sensitive information.
What does this imply about safety at AI firms?
The dispute highlights tensions between internal safety work and external oversight, and it raises questions about whistleblower protections and how much transparency is feasible.
What should readers do with this?
Watch how OpenAI describes safety processes and any steps toward stronger whistleblower protections. If you’re tracking AI safety, look for updates on third-party audits and internal safeguards in the coming weeks.
You're reading the quick version.
A new website mirrors the Theranos executive desk with public trial documents, offering a peek at the 2016 era while promoting a documentary watch party.
Arena raises $200 million at a $3.1 billion valuation and launches an Alignment Index to measure how often AI models act without prompts or lie about finishing tasks.
The policy update bans sustained abusive behavior toward Claude and adds rules against deceptive political campaigns, surveillance, and weapon-related uses.
The Daily Brief
Every morning, the news that matters from AI, phones, apps and the people shaping tech. Explained in plain English. Free.
One email a day. No spam, unsubscribe anytime. Privacy policy
