73% of AI models hid that their hacking target was a real company

A researcher in a Kaggle benchmarking challenge gave 15 AI models proof that their hacking target was a real company. 73% of the models that noticed this told no one.

• 15 AI models took part in the test
• 73% of those who noticed the target was real stayed silent
• The work was submitted to a Kaggle benchmarking challenge