AI news · October 10, 2026
Three fired OpenAI researchers ask for transparency about their firings
Three former OpenAI researchers say their firings followed safety work and ask for clearer rules; the dispute matters because model monitoring depends on access to independent checks, not only lab statements.
- researchers in the letter
- 3
The Rundown and The Verge covered an Oct 8 open letter from Tomek Korbak, Jasmine Wang, and Mikita Balesni. The researchers dispute OpenAI's allegation that they shared confidential information with an outside safety group and say they acted in line with OpenAI's mission. OpenAI has described the firings as misconduct and told reporters it will clarify collaboration rules, outside review, and model monitoring.
The dispute follows the companies' work with METR, the safety group that investigated an earlier model incident. The public accounts do not resolve the facts, and the letter is not an independent finding. For builders, the useful issue is access: a safety check needs reproducible inputs, logs, and enough independence to be rerun when a model, tool, or policy changes.
What you can do with it
If you run an AI product, separate model development from evaluation access. Keep reproducible prompts, logs, and a named reviewer; make it possible to rerun a safety check when a model or tool changes. This is a process lesson, not a claim that your team has OpenAI's facts.
Our take
The important part is not choosing a side from two public accounts; it is that monitoring fails when evaluators lose access to the system they review. The catch is incomplete evidence: the letter and company statements describe the dispute, but neither is an independent ruling.
Links The Rundown report
Source: The Rundown AI — Made With Models writes the brief; the reporting is theirs.