chains

A third AI-related proof-of-concept attack that garnered attention used a prompt injection to cause GitLab’s Duo chatbot to add malicious lines to an otherwise legitimate code package. A variation of the attack successfully exfiltrated sensitive…