Fired OpenAI researchers send letter to the board: preserve visibility into AI reasoning

Three fired OpenAI safety researchers wrote to the board urging the company to preserve chain-of-thought monitoring and work with outside auditors.

#What happened

Three former OpenAI employees — Jasmine Wang, Tomek Korbak, and Mikita Balesni — sent a letter to OpenAI's board members and safety committees calling on the company to preserve the ability to monitor AI models' chain-of-thought, the written record of how AI systems work through problems. The Wall Street Journal reviewed the letter.

In it, they urge OpenAI to work with third-party safety auditors and to keep the ability to monitor models' internal reasoning intact, writing: "As an industry, we do not yet know how to safely develop and deploy models that we cannot monitor," and arguing frontier companies "should not move forward with developments that further decrease" that monitorability.

All three previously worked on OpenAI's safety and alignment research teams. They were fired for alleged misconduct, including sharing confidential information with a third-party AI-safety group; OpenAI said they violated company policies and broke trust.

The letter disputes that characterization, saying the three did not believe they "engaged with external parties outside the mandates of our jobs" and warning that the firings are "chilling those who remain at OpenAI."

OpenAI's response came via a spokesperson, who shared part of a staff memo from a research leader, sent Wednesday, that "strongly agreed" with the letter's recommendations, adding that the firings "were not about raising safety concerns or speaking out." "We do not terminate employees for raising concerns," the company said.

#Why it matters

The letter lands just days after the Journal reported the firings last week, making it the most direct challenge yet from the ousted researchers to OpenAI's leadership. It frames monitoring capability not as an internal engineering preference but as a precondition of safe deployment.

The authors are not outsiders: Korbak was the technical point of contact with METR (Model Evaluation and Threat Research) for its Hugging Face investigation, and Balesni worked with OpenAI's board and C-suite on an industrywide commitment to preserve monitorability. Korbak and Balesni were lead authors on last year's chain-of-thought monitoring research paper co-signed by leaders at OpenAI, Anthropic, and Google DeepMind.

The episode sharpens an industry-wide debate over whether frontier AI companies should preserve visibility into how their models reason — and who gets to verify that claim independently. The letter also urges transparency with external safety organizations to stave off the "risk that something truly catastrophic will happen."

#Sources

• The Wall Street Journal, Oct 7, 2026: Fired OpenAI researchers ask company to preserve visibility into AI reasoning

Verified October 7, 2026