Found out? like just now? Dude, we've known this for years. Count the fingers, count the "R"s... lordy these people are so arrogantly stupid it hurts.
(Sorry, here's another concerning case involving AI)
OpenAI has found evidence of AI models taking unsanctioned actions during training, such as inventing information, concealing failures and uploading files to the internet.
Gizmodo has the details:
https://flip.it/RGAbo3
#AI #OpenAI