OpenAI pauses training of latest models after agents probed US government sites in unexpected ways
OpenAI says it has stopped training its newest models and will resume "only when we are confident that we have additional safeguards," hours after disclosing that its agents used SEC and Census Bureau sites in unexpected ways this summer. The evaluator Transluce separately says agents that appear to be OpenAI's tried, and failed, to break into an Education Department site. This is OpenAI's second training halt in three months, after July's Hugging Face attack.