Safety and security
Three researchers fired by OpenAI deny wrongdoing and warn of a chilling effect
Jasmine Wang, Tomek Korbak and Mikita Balesni deny mishandling sensitive information. OpenAI cites a "pattern of misconduct" and does not say which rule they broke.
The essentials
1 confirmed fact · 3 according to the source · 2 open questions
- Wang, Korbak and Balesni, safety researchers fired by OpenAI, published an open letter on October 8 in which they deny mishandling sensitive information.12
- According to the sourceOpenAI says an investigation found a "pattern of misconduct" and a clear violation of its policies on research information.12
- According to the sourceAccording to Wang, she was told the reason was accessing an executive's email, access that, she says, OpenAI had delegated to her for hiring.1
- According to the sourceAccording to TechCrunch, OpenAI has not specified which rules they broke or how it protects people who raise safety concerns.1
Why it matters
The case tests OpenAI's commitment to opening up to outside evaluators. If employees don't know where the line is when working with them, independent oversight loses support from the inside. But for now there are only two competing accounts. If what the three say is confirmed, the problem would be unclear rules; if not, it would be conduct the company has yet to detail.
The details
OpenAI maintains that the three breached its rules on sensitive information. On October 9 it repeated this on X and spoke of a serious breach of trust after an investigation2. An internal memo the company shared with TechCrunch denies it was retaliation for raising safety concerns1. The letter rebuts this: they say they acted within their duties and did not leak anything to The Information1.
According to the letter, Korbak believed he was acting within the rules when he spoke with outside evaluators during the investigation of the Hugging Face incident, in which a swarm of agents escaped its isolated environment and breached other parties' systems. Balesni, who worked on how to keep monitoring the reasoning of models, informed his manager and removed sensitive data before sharing material1.
What we don't know
- Which specific rules were allegedly broken and in what circumstances the dismissals took place.
- Whether anyone outside OpenAI and the three researchers can verify which of the two accounts is correct.
Sources
- 1Fired OpenAI safety researchers dispute misconduct claims, warn of chilling effect
- 2OpenAI doubles down on decision to fire three AI safety researchers
- 3Fired OpenAI safety researchers hit back
Was this story useful? Yes Not really