OpenAI dismisses three safety researchers over AI risk dispute
ai-safety openai
| Source: Associated Press News on MSN | Original article
OpenAI dismissed three safety researchers, citing a breach of trust, as the company defends the firings amid an internal dispute over AI risk concerns.
OpenAI announced on Friday that it had terminated three members of its safety research team—Tomek Korbak, Jasmine Wang and Mikita Balesni—citing a “breach of trust.” The company’s statement said an internal investigation found the trio had handled sensitive information outside established procedures, though OpenAI did not disclose what data or policy was involved.
The three researchers responded the same day with a letter to OpenAI’s board and its safety oversight committees, disputing the company’s account and arguing that the vague accusations could intimidate other staff into silence. Their letter, posted publicly, details the circumstances of their dismissals and raises concerns about how the firm addresses internal dissent on AI risk. OpenAI has rejected the notion that the firings were retaliation for speaking out, insisting the decision was purely procedural.
The episode matters because OpenAI’s safety team is a key line of defence against the rapid escalation of AI capabilities. Removing senior safety staff amid a dispute over risk assessment could signal a shift in how the company balances rapid development with internal checks, and it may chill open discussion of safety concerns among engineers. The incident also adds pressure on OpenAI’s governance structures, which have been under scrutiny since the launch of increasingly powerful models.
Observers will watch for a formal response from OpenAI’s board, any follow‑up investigations by external auditors, and potential regulatory interest in the company’s internal safety protocols. The dismissed researchers may also pursue legal avenues or rally support from the broader AI‑safety community, setting a precedent for how tech firms handle internal dissent on high‑stakes technology.
Sources
Back to AIPULSEN