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.

Read the full story at Decrypt →