Openai Stands By Firings As Safety Researchers Dispute Its Account
OpenAI says three researchers breached rules for sensitive information; the former employees say their dismissals could discourage safety discussions.
OpenAI is standing by its dismissal of safety researchers Jasmine Wang, Tomek Korbak and Mikita Balesni, saying an internal investigation found a significant breach of trust. The three dispute the company’s account. In an open letter published Thursday, they said they had followed OpenAI’s mission and the working norms in place at the time, and warned that the handling of their firings could make other employees afraid to speak up.
Table of Contents
OpenAI’s explanation
In a Friday post on X, OpenAI said the researchers were dismissed for violating policies on handling sensitive information, according to The Verge. The company said the decision was not a response to their criticism of OpenAI or their concerns about AI safety. It also said its investigation found breaches beyond those addressed in the researchers’ letter, without providing details.
Engadget reported that the three were dismissed last week after allegedly sharing information with an external AI safety organization. OpenAI’s earlier statement cited its policies on accessing and handling sensitive company information and said the employees had broken the trust required for their work.
The researchers’ response
Wang, Korbak and Balesni said OpenAI’s internal and public communications about their dismissal risk discouraging colleagues from raising concerns or working with outside safety experts. They described open disagreement and consultation with independent safety organizations as established parts of their work at OpenAI, and said employees may now be unsure which practices are permitted.
According to Engadget, the former employees said they did not believe their dealings with external parties went beyond their mandate. They said those communications took place in coordination and discussion with board members and senior executives. They also denied being the source of a leak to The Information about OpenAI architectures. Wang said she had alerted an executive after accidentally clicking on a sensitive email.
Questions about safety collaboration
The researchers urged OpenAI to maintain its public commitments to bring in third-party safety auditors and not treat their firing as a reason to retreat from those partnerships, Engadget reported. They also called for frontier models to remain monitorable and for continued dialogue between OpenAI researchers and the broader safety community.
The disagreement leaves a central question unresolved in the public accounts: OpenAI says its investigation established policy violations, while the researchers say their conduct fit the company’s working norms. OpenAI has not publicly detailed the additional breaches it says it found.
Sources
This story was compiled by AI from the reports below. Read the originals for the full details.