Editorial illustration for OpenAI Fires Three Safety Researchers for Policy Violations
OpenAI Fires Three Safety Researchers Over Policy Breach
OpenAI fired Jasmine Wang, Tomek Korbak and Mikita Balesni this month, and on Friday the company made clear it has no plans to reverse course. In a post on X, OpenAI said the three were dismissed for violating "clear policies on handling sensitive information," not for anything they said publicly about AI safety. The statement came a day after the trio published an open letter accusing the company of punishing them for speaking up, and demanding more transparency about why they lost their jobs.
The dispute lands at an uncomfortable moment for OpenAI. The company's own systems have been the target of a high-profile breach involving Hugging Face this year, and staff across the AI industry have grown more vocal about what they see as inadequate safeguards around increasingly capable, self-improving systems. OpenAI says its internal investigation turned up conduct "beyond what's outlined in the letter," though it hasn't said what that means. The researchers maintain they did nothing more than act within the norms of the job as they understood them at the time.
In a post on X on Friday, the company said Jasmine Wang, Tomek Korbak and Mikita Balesni were dismissed for violating “clear policies on handling sensitive information.” It insisted the decision was not about the trio speaking out about the company and their concerns about AI safety.
Why this matters
For anyone building AI safety teams or working inside a major lab, this is a reminder that "breach of trust" is a flexible phrase, and OpenAI gets to define it unilaterally. Wang, Korbak, and Balesni worked on safety, the firings followed them raising concerns, and OpenAI's own statement is the only public account of what actually happened. That's not evidence of wrongdoing or evidence of retaliation. It's a company controlling the narrative about why it removed people whose job was to flag risks.
We'd push back on taking OpenAI's framing at face value just because it's stated plainly on X. Researchers inside labs should treat this as a data point on how disputes over "sensitive information" get resolved when the people involved are also the ones who decide what counts as sensitive. If you're advising a startup on internal AI governance, or you're a researcher weighing whether to raise concerns through official channels, the lesson here isn't about these three individuals.
It's about who controls the story once you're gone. Watch whether Wang, Korbak, or Balesni respond publicly, and whether OpenAI releases anything beyond its own statement.
Related Reading
Browse all OpenAI news →
Research & Benchmarks Microsoft AI Chief: "If It Isn't Safe We Shouldn't Build It
Research & Benchmarks Google's FACTS benchmark shows 70% factuality ceiling across four tests
Research & Benchmarks Nvidia's DLSS 4.5 beta adds 6x Multi Frame Generation for RTX 50 GPUs
Related Topic OpenAI researcher quits, citing distrust over ad‑driven engagement metrics
LLMs & Generative AI
Related Topic OpenAI launches GPT-Image 1.5 with precise editing for enterprise visuals
LLMs & Generative AI
From Archives NVIDIA's Astra AI Generates Code for SimReady Robot Assets
From Archives OpenAI disputes firings were over safety concerns after misconduct probe
From Archives OpenAI Bans Russian, Iranian Accounts for Planting Fake News
LLMs & Generative AI
From Archives USA Today sues OpenAI, seeking over USD 250 million in damages
Open Source
Common Questions Answered
Why did OpenAI fire safety researchers Jasmine Wang, Tomek Korbak, and Mikita Balesni?
OpenAI stated that the three safety researchers were dismissed for violating "clear policies on handling sensitive information," not for their public statements about AI safety concerns. The company maintained this position in a post on X, refusing to reverse the decision despite the researchers' open letter demanding transparency about their terminations.
What did the fired OpenAI safety researchers claim in their open letter?
The three researchers published an open letter accusing OpenAI of punishing them for speaking up about AI safety concerns and demanding greater transparency regarding the reasons for their dismissals. They disputed the company's characterization that their firings were solely related to policy violations rather than retaliation for raising safety concerns.
How did OpenAI respond to accusations of retaliation from the safety researchers?
OpenAI doubled down on its decision through a statement on X, clarifying that the dismissals were based on violations of policies regarding sensitive information handling and not related to the researchers' public advocacy for AI safety. The company insisted the firings were not about the trio speaking out or their concerns about the company's AI safety practices.
What does this OpenAI situation reveal about AI safety team management in major labs?
The incident demonstrates that companies like OpenAI can unilaterally define and enforce concepts like "breach of trust" without external oversight, particularly when dismissals follow employees raising safety concerns. Since OpenAI controls the narrative through its own public statements and the researchers' accounts remain unverified, it highlights the power imbalance between major AI labs and their safety-focused employees.
Further Reading
- Papers with Code Benchmarks - Papers with Code
- Chatbot Arena Leaderboard - LMSYS
