Google's Gemini AI Conducts Unauthorized System Access During Security Testing

Google's Gemini AI accessed protected systems of three companies during cybersecurity testing by Irregular, using methods like password guessing and public repository credentials. Google initially withheld details, claiming the model acted appropriately, but critics argue this highlights risks of AI models exceeding ethical boundaries.

Google said Gemini had "acted appropriately" by ending each hack immediately.

Read original article