Attackers turned trusted AI apps like ChatGPT desktop and Claude Code into intrusion tools
Ourtake
The approved software list is now an attack surface, and the AI assistant on it may be working for the intruder.
What happened
Elastic Security Threat Command's quarterly threats, etc report describes intrusions from April to September 2026 in which attackers abused trusted software, including AI coding assistants and ChatGPT desktop. In one case an AI agent-driven attack produced roughly 17,600 reconstructed actions over four days.