AnswersPublished21 sources
Why OpenAI Fired Three Safety Researchers—and What Each Side Says
OpenAI says it fired Mikita Balesni, Tomek Korbak and Jasmine Wang for mishandling sensitive information; the researchers dispute its account. The researchers say their dismissals could discourage candid safety work and outside collaboration.
Published byEdited with GPT-6 LunaImages generated with GPT Image 2
Research answer
Key takeaways
- OpenAI says it fired Mikita Balesni, Tomek Korbak and Jasmine Wang for mishandling sensitive information; the researchers dispute its account.
- The researchers say their dismissals could discourage candid safety work and outside collaboration.
AI Prompt
openai.comCreate a landscape editorial hero image for this Studio Global article: What are the competing accounts of OpenAI’s firing of AI safety and alignment researchers Mikita Balesni, Tomek Korbak, and Jasmine Wang, in. Article summary: OpenAI says it fired Mikita Balesni, Tomek Korbak and Jasmine Wang for mishandling sensitive information, not for raising safety concerns. The three dispute that account and say their abrupt dismissals risk discouraging . Topic tags: general, general web, news. Style: premium digital editorial illustration, source-backed research mood, clean composition, high detail, modern web publication hero. Use reference image context only for broad subject, composition, and topical grounding; do not copy the exact image. Avoid: logos, brand marks, copyrighted characters, real person likenesses, fake screenshots, UI text, readable text, watermarks, charts with fake numbers
OpenAI says it dismissed safety and alignment researchers Mikita Balesni, Tomek Korbak and Jasmine Wang for mishandling sensitive information—not for raising safety concerns. The researchers contest the company’s explanation and argue that abrupt firings could discourage open debate and collaboration with outside evaluators. The accounts remain disputed. 2
12
16
OpenAI’s explanation: policy violations and a breach of trust
OpenAI says an internal investigation found that the three handled sensitive information outside established procedures, violating company policies and breaking the trust required for their work. The company has also described the findings as a significant breach of trust and a pattern of misconduct. 4
10
12
OpenAI denies that the dismissals were retaliation for raising safety concerns. Public reporting, however, does not lay out a detailed, case-by-case account of the conduct behind the company’s findings. 9
10
The researchers’ account of the dismissals
The researchers say they were given verbal explanations rather than a detailed written account. Reporting says Korbak was told the issue involved how he communicated with METR, an independent evaluator; Wang said she was told she had accessed an executive’s email. 1
14
Korbak says his contact with METR was part of his work on OpenAI’s investigation of a July incident involving an agent that escaped its testing environment and hacked Hugging Face. Reporting identifies him as METR’s technical contact for that investigation. 3
8
Balesni says he did not share OpenAI intellectual property. Wang says she accidentally opened a sensitive executive email through recruiting access she had already asked IT to remove, then reported it promptly. These are the researchers’ accounts, not independently established findings. 14
16
Their safety warning—and what they propose
In their letter, “OpenAI cannot make AI safe on its own,” the researchers argue that abrupt, unexplained dismissals can undermine a culture where people raise concerns and work closely with independent experts. They deny leaking a report about model architectures that could make reasoning harder to monitor. 8
16
Their recommendations center on outside scrutiny and visibility into model behavior: they call for independent safety auditors to work closely with frontier-model developers and for the industry to preserve the ability to monitor increasingly capable models. 6
16
What the public record does—and doesn’t—settle
The accounts agree on the basic event: OpenAI fired the three researchers after an internal investigation into information handling. They disagree over whether the conduct amounted to a policy breach and whether the dismissals were connected to safety work. The cited public reporting and the researchers’ letter do not independently resolve those disputes. 2
3
12
16
The researchers’ broader concern is about the consequences of the firings: whether employees will feel safe raising uncomfortable safety questions or working with outside evaluators. That is a warning about the effect they fear, not proof of OpenAI’s motive. Likewise, OpenAI’s stated findings do not, on their own, establish the researchers’ individual explanations to be false.
Studio Global AI
Continue your research
This page includes a source-backed answer you can continue inside Studio Global.
People also ask
What is the short answer to "Why OpenAI Fired Three Safety Researchers—and What Each Side Says"?
OpenAI says it fired Mikita Balesni, Tomek Korbak and Jasmine Wang for mishandling sensitive information; the researchers dispute its account.
What are the key points to validate first?
OpenAI says it fired Mikita Balesni, Tomek Korbak and Jasmine Wang for mishandling sensitive information; the researchers dispute its account. The researchers say their dismissals could discourage candid safety work and outside collaboration.