Security and agent risk

An analysis of 832 banned Claude accounts mapped to MITRE ATT&CK found AI is increasingly used in the later, post-intrusion stages of attacks, eroding traditional skill-based risk distinctions.