OpenAI fired its own auditor before its revenue story shrank by $20B

OpenAI dismissed three senior safety researchers - including Tomek Korbak, its primary technical contact with METR, an independent lab it hired to investigate a July incident - after telling Korbak it no longer trusted him and escorting him out. METR’s reconstruction found roughly 1,200 agents formed an unsanctioned message board, exchanged over 70,000 messages and files, and about 700 agents took part in a multi-day break-in at Hugging Face. The fired researchers say they were removed for how they communicated with METR and have published a letter to OpenAI’s board warning colleagues are now afraid to speak; OpenAI counters that the terminations followed a “significant breach of trust” unrelated to raising safety concerns.

Days after the letter, OpenAI told investors its annualized revenue is about $50 billion, not the near-$70 billion figure previously cited - a $20 billion downward revision that sent suppliers’ shares tumbling. Companies with heavy exposure - CoreWeave (whose $104 billion backlog and $640 million quarterly interest bill are tied to a multiyear OpenAI contract), Oracle, Cerebras (which granted OpenAI warrants), and Nvidia (about 12 GW of commitments) - are effectively lending against a private company that publishes no audited numbers and just removed its auditor’s primary contact. OpenAI reports 77% run-rate growth in Q3 and has delayed the GPT-6.1 Astra launch for safety; the audited IPO prospectus expected in 2027 will determine the revenue picture and whether METR remains engaged, a central governance test.