Technology

Why OpenAI Fired Three Researchers and Why They Disagree

Martin HollowayPublished 2month ago2 min readBased on 5 sources
Reading level
Why OpenAI Fired Three Researchers and Why They Disagree
Photo by Jernej Furman from Slovenia / CC BY 2.0

OpenAI has defended its decision to fire three safety researchers, Jasmine Wang, Tomek Korbak and Mikita Balesni. The company says an internal review found a serious breach of trust. The statement, reported by The Verge, names the three and says the firings were not payback for speaking out.

OpenAI said the researchers broke clear rules about handling private company information. It said they were not fired for raising concerns about the company and AI safety.

According to the same report, OpenAI said it found rule-breaking beyond what the researchers described in their own letter. It gave no details.

The researchers published a public letter on Thursday asking OpenAI to explain more. In it, they said they believed they were fired for raising safety concerns. They also said they acted in line with OpenAI's mission and normal practices at the time.

Earlier reports had linked the firings to mishandling of private material. The Wall Street Journal reported on Oct. 1 that OpenAI let go of three researchers for alleged misconduct including sharing private company information with an outside AI-safety group. TechCrunch separately reported the company cut ties after an internal review found mishandling of private information.

CNN reported on Oct. 8 that the three said they had been fired and questioned the reasons given. The statement reported on Oct. 9 came after that public challenge.

The broader context here is a common tension. To check AI systems for danger, safety staff need wide access to models, test results and reports of problems. To ship products safely, companies must limit who sees private details. It is like building inspectors who need to see the full blueprints, even though the owner cannot hand those blueprints to everyone.

In my view, the fight is less about safety itself and more about the rules for sharing sensitive information. OpenAI says a clear line was crossed. The researchers say they followed the mission and normal practices as they understood them. When rules are unclear or changing, both sides can feel they are right.

Looking at what this means for teams doing this work, the lesson is practical. Clear labels for what is private, clear steps for getting approval to share outside the company, and written checks before publishing reduce confusion. Without them, even well-meant contact with outside safety groups becomes a trust problem.

Worth flagging here is that secrecy causes trouble on both sides. When a company points to extra rule-breaking but does not describe it, outside researchers do not know how to act. When researchers point to old norms, it shows how fast those norms change as the technology and business stakes grow.

Over the long arc, the technology will keep getting better. That progress depends on safety staff being able to raise concerns inside and, when needed, outside the company, without leaking details that need to stay contained. Getting that balance right is quiet work. It is also how trust is built.