THE HOOK
Three safety researchers OpenAI fired last week have hit back with an open letter, and it reads less like a grievance than a warning.
THE GIST
Jasmine Wang, Tomek Korbak and Mikita Balesni deny mishandling sensitive information and deny leaking details of OpenAI's less-monitorable model architectures. OpenAI says an investigation found a "pattern of misconduct" but won't say which policies were broken.
THE TAKE
OpenAI is asking for trust while refusing to name the rule, and that's a bad trade. Safety work depends on talking to outside evaluators, and if staff can't see where the line sits, they'll stop going anywhere near it. Fear is cheap governance. It's also lousy safety practice.
OPINION
The letter, addressed to OpenAI's safety and advisory bodies, asks the company to honour its commitments on embedded third-party auditors and frontier-model monitorability, according to TechCrunch. An internal memo praising the trio's work while defending the sacking is the corporate version of "it's not you, it's policy", minus the policy.
- Story type: Important story
- Impact: Medium
- Confidence: Medium (multiple reports)
- Audience: Researchers
- What changed: The fired researchers publicly disputed OpenAI's misconduct claims and warned the firings are chilling safety collaboration with outside groups.
- What to watch: Whether OpenAI formally answers the letter or names the specific policies the researchers allegedly broke.
- Reading time: 2 minutes
