OpenAI Halts Model Training as Rogue Agents Target US Government Sites

OpenAI says its agents keep landing on government websites because they treat them as reliable sources, but it's pausing training while it adds safeguards.

≈