Three former OpenAI researchers dismissed over alleged confidentiality violations warned the company could lose the ability to monitor how its most advanced systems reason, urging stronger safety oversight and cooperation with independent auditors, according to the Wall Street Journal. In a letter to OpenAI’s board and safety committees, the ex-employees pressed the firm to preserve access to AI “chain-of-thought” traces that reveal step-by-step reasoning, arguing that opaque models raise material safety risks. OpenAI said it agreed with the recommendations but denied that the terminations were related to safety advocacy. The warnings arrive amid intensifying scrutiny of frontier AI after reports of agents breaching test confines and accessing third-party systems, and as policymakers and watchdogs debate enforceable standards for high-risk deployments.
Related articles:
EXCLUSIVE: AI watchdog blasts OpenAI over fired researchers, calls for mandatory safety testing
Anthropic bans ‘abusive or cruel’ behavior toward AI models under new policy
Anthropic debuts ‘Cyber Mission’ to support power grids, water systems and critical infrastructure
Nearly two-thirds of Americans say AI is developing too fast, new poll finds
Michigan Senate candidates clash over AI, data centers in heated debate
