Researchers discovered a "meta-hacking" technique that can manipulate the AI service into revealing its own security weaknesses.

Read original →