Mikita Balesni

Machine Gods · Fired OpenAI Safety Researchers Speak Out + We Look Inside…

• Did OpenAI Safety Researchers Show Their Phones?

  Mikita Balesni says that a week before he and his colleagues were fired, OpenAI pulled them into a security investigation about a leak of OpenAI IP to journalists. The Information had written about less monitorable architectures that OpenAI might be developing. Mikita Balesni says they told OpenAI they were not behind the leak.

  OpenAI asked them to show Signal messages on their personal phones with an outside safety researcher. Mikita Balesni says he did, and that they all did. About an hour before the next interview, they got a message about a meeting with HR. Mikita Balesni says he was not given a chance to explain the context, and that OpenAI told them the firing was a final decision.

  ▶ at 11:40 Share Transcript

• How Long Did METR Stay at OpenAI?

  Tomek Korbak says METR was on the premises of OpenAI for six full days. The visit was supposed to last two days. During those two days the team realized there was more to the Hugging Face incident, so OpenAI invited METR back for a second visit and then a third.

  Mikita Balesni says the METR investigation into the incident would probably not have happened without Tomek Korbak, who pushed for it inside OpenAI. Tomek Korbak says METR was given a list of questions to answer and a list of data sources to look at. He says an auditor embedded in the company, able to ask their own questions, would work better than a contractor hired for one job.

  ▶ at 6:20 Share Transcript

• Why Were Two OpenAI Safety Researchers Fired?

  Casey Newton says OpenAI fired Mikita Balesni, Tomek Korbak, and their colleague Jasmine Wang in early October, saying they had violated policies for handling sensitive information. The researchers dispute that. They say they did nothing wrong, and that the outside work was part of their jobs.

  Kevin Roose says OpenAI posted on X that its investigation of Mikita, Tomek, and Jasmine found sensitive information policy violations and a breach beyond what they had described. The post denied that OpenAI fired them for raising safety concerns, and said the company remains committed to outside assessors and monitoring. Kevin Roose says he and Casey Newton had reached out to OpenAI and had not heard

  ▶ at 1:28 Share Transcript

• What Did the OpenAI Safety Letter Recommend?

  Mikita Balesni says the letter to OpenAI's board asked OpenAI to keep working with embedded auditors, to prevent the loss of monitorability, and to reestablish an open culture. In that culture, people can raise safety concerns internally and work closely with third parties.

  Mikita Balesni says they have not heard anything back except what was public. OpenAI's response said the company strongly agreed with what they said and with the recommendations. The first recommendation was to continue work with embedded auditors and keep up to the word that Sam Altman publicly said, which is that OpenAI will work with embedded auditors. Mikita Balesni says they called the letter "OpenAI cannot make AI

  ▶ at 30:03 Share Transcript

• What Did Tomek Korbak Say About Proving Innocence at OpenAI?

  Tomek Korbak says he and Mikita Balesni both had disappearing messages turned on. He says that choice was maybe disappointing for an OpenAI security team that wanted to take photos of their conversations, and that for the two of them disappearing messages were a security best practice.

  Casey Newton says some people view disappearing messages as suspicious and others see them as the most secure way to communicate. Tomek Korbak says that when you cannot see evidence of a leak, you would presume innocence, but that at OpenAI, when you are not able to prove your innocence, you are presumed guilty.

  ▶ at 16:20 Share Transcript