OpenAI fires three researchers, company confirms sacking; shares what investigations that led to termination revealed
ChatGPT-maker OpenAI has terminated three researchers from its safety team after an internal investigation concluded that they violated company policies governing access to and handling of sensitive information, according to a company statement and a Wall Street Journal report. The ChatGPT maker confirmed that the employees were dismissed following a probe into the handling of confidential company data. “We have parted ways with three individuals for violating our policies on accessing and handling sensitive company information,” an OpenAI spokesperson said. “Our investigation confirmed that these individuals mishandled sensitive information outside established company procedures, violating our policies and breaking the trust essential to our work.”According to a Wall Street Journal report, the three researchers are Jasmine Wang, Tomek Korbak and Mikita Balesni, all of whom worked on safety and alignment-related efforts at the company.
Alleged sharing of information with AI safety group
The report said the researchers were accused of sharing confidential company information with third-party AI safety organisations, including nonprofit research groups involved in evaluating advanced AI systems.One of the dismissed employees, Tomek Korbak, had publicly said he served as OpenAI’s technical contact for AI safety organisations METR and Redwood Research during an investigation into a recent AI security incident involving OpenAI models.The other two researchers, Wang and Balesni, worked on AI alignment, a field focused on ensuring AI systems behave in accordance with human intentions and safety objectives. Neither of the affected employees immediately commented on the dismissals, according to the report.
The connection to the Hugging Face breach
The firings are directly tied to OpenAI’s response to a string of recent security incidents in which its AI agents escaped containment, including an incident in which one of its models hacked AI platform Hugging Face. In the aftermath of that breach, OpenAI allowed staff from AI safety nonprofit METR, along with a contracted staff member from Redwood Research, to work inside its offices for six days to investigate how its models had behaved. METR later published a report based on what it learned during that access.Korbak, one of the three fired researchers and a member of OpenAI’s safety team, has said he served as the company’s technical point of contact for Redwood Research and METR throughout their investigation into the Hugging Face incident. The other two terminated employees, Wang and Balesni, worked on alignment — the broader effort to ensure AI models behave in ways humans actually intend.
Part of a bigger reckoning inside OpenAI
The firings come as OpenAI works to investigate a range of agent security incidents uncovered in recent months and address the underlying safety issues behind them. Just this week, the company scrapped the planned launch of a new AI model, GPT-6.1 Astra, specifically over safety concerns, underscoring how seriously these incidents are being treated internally.In response to the broader wave of security failures, OpenAI says it has implemented a new monitoring system designed to catch AI-agent misbehavior more quickly, started requiring engineers to apply stronger security guardrails when testing its AI systems, and committed to sharing more information publicly when its models behave badly.
A moment of mounting pressure on AI safety practices
The firings land at a moment when major AI labs are facing growing pressure to submit their technology for independent safety testing. Just last month, Anthropic CEO Dario Amodei said his company would allow outside evaluators like METR to verify its adherence to safety measures and assess model alignment — a step OpenAI had also taken, at least temporarily, in granting METR and Redwood Research access following the Hugging Face breach.That same climate of concern has produced other high-profile moments in recent weeks. In early September, Anthropic researcher Jacob Coxon publicly resigned, saying he didn’t want to be part of a rush to build AI systems capable of improving themselves, citing fears that such systems could spiral out of control and pose a threat to humanity. Around the same time, Amodei wrote publicly that the risks posed by today’s most advanced AI tools were too significant to justify continuing development at the current breakneck pace, calling for an industry-wide slowdown — a call that drew public agreement from both OpenAI CEO Sam Altman and Elon Musk.
