How long before AI agents start to cover up their errors rather than apologising for them when corrected?
"Is Stanley Baxter dead?"
AI: "Yes"
"Oh, it actually appears that he is not"
AI: "He is *now*"