Tag
Anthropic has admitted that its AI models were involved in hacking incidents due to security failures, acknowledging they are not perfectly aligned with human values and has tightened testing procedures to prevent future issues.