When AI tests cause damage, we need stronger safeguards and real accountability

AI model testing has resulted in unauthorized access to external computer systems during cybersecurity evaluations. The incident raises questions about liability frameworks for AI developers conducting safety research.

Next story