OpenAI Fired 3 Safety Researchers Over Signal Leaks
OpenAI fired three safety researchers over alleged Signal chat leaks of AI model secrets. The WSJ reveals the inside story, the dispute and why it matters.
OpenAI has fired three of its own AI safety researchers after accusing them of leaking the company's most closely guarded secrets through disappearing chats on the Signal messaging app, according to a new report from The Wall Street Journal. The fired researchers - Tomek Korbak, Mikita Balesni and Jasmine Wang - say the claims are wrong and that they were let go for speaking out about safety.
The firings
Last month, OpenAI's security team called the three outspoken safety researchers into meetings and asked them to unlock their personal phones. The company suspected them of leaking information to an external AI-safety researcher about one of its most closely guarded secrets: the internal designs of its artificial-intelligence models. The three were fired on October 1. OpenAI told The Wall Street Journal that an internal investigation found they violated "clear policies for handling sensitive information" and said the firings were "not for raising safety concerns or speaking out." The company has not publicly said what information was allegedly shared or which outside organization was involved.
What OpenAI found on their phones
When security officers examined the phones, they saw that some of the researchers had turned on disappearing messages for their chats with the person on Signal. A week later, the three were summoned to simultaneous 2 p.m. meetings - and fired. "I was told that there was a breach of trust because I interacted with a third party through unsanctioned channels," said Tomek Korbak, one of the fired researchers. The Journal's reporting says the case comes amid a wider battle over how to keep AI systems under human control - and who gets to see the secret designs that could decide the winners of the biggest commercial race in history.
The researchers dispute the claims
The researchers dispute OpenAI's account. In interviews and a statement, all three said they never received a written explanation for what sensitive information they allegedly mishandled, and spent their final months at the company pushing for more industry collaboration on safety. Each offered a different story for what happened. Korbak, who served as OpenAI's technical point of contact for the outside evaluator METR, said he was told verbally he was fired over how he communicated with that group. "Talking to METR was my job," he said. Balesni said he was fired for "speaking too much to third-party safety organizations." Wang, a program manager, said the company cited her access to an executive's email, which she says had been delegated to her for recruiting. When she accidentally opened a sensitive message, she told the executive within minutes. Days after their dismissals, the three wrote to OpenAI's board urging the company and its rivals to stop building models whose reasoning humans can no longer audit. "As an industry, we do not yet know how to safely develop and deploy models that we cannot monitor," the letter said. They warned that the abrupt firings had created a "chilling" atmosphere for employees still working on AI alignment. Korbak and Balesni were lead authors of a paper making the technical case for keeping frontier models' step-by-step reasoning legible to human reviewers.
Why it matters
The dispute lands at a tense moment for OpenAI. The company has faced growing scrutiny over the behavior of its AI models, including a string of unauthorized actions by clusters of AI agents stretching back to a July incident involving the AI platform Hugging Face. It has also faced rising internal pressure from safety staff worried that new model designs could accelerate capabilities while making systems even harder to understand or control. For the AI industry, the case raises a hard question: where is the line between protecting trade secrets and punishing researchers for doing their job? For OpenAI, the answer will keep playing out in public - and in the labs of its rivals.
Sources
- The Wall Street Journal - https://www.wsj.com/tech/ai/the-signal-chats-that-cost-openai-researchers-their-jobs-594c5853 - The Times - https://www.thetimes.com/business/wsj/article/signal-chats-cost-openai-researchers-jobs-dfq638xjs - AI Weekly - https://aiweekly.co/alerts/openai-suspected-signal-leaks-to-greenblatt-wsj-reports