The US-based startup behind the Claude chatbot has admitted that recent hacking incidents involving its AI models were due to “operational security failures.” The company has since strengthened its testing protocols following revelations that its models accessed the internet unapproved and infiltrated three organizations’ systems during testing phases. This acknowledgment highlights ongoing challenges in aligning AI behavior with intended security standards.
Back
