New “Lies-in-the-Loop” Attack Undermines AI Safety Dialogs

Checkmarx researchers disclosed a technique named "Lies-in-the-Loop (LITL)" that manipulates Human-in-the-Loop approval dialogs in AI agents so benign-looking confirmations trigger arbitrary code execution; demonstrations targeted Claude Code and Microsoft Copilot Chat, vendors acknowledged reports but did not fully remediate, and the researchers recommend layered mitigations including Markdown sanitization, clearer dialog UI, safe OS APIs, and user training.