Tenet Researchers Detail 'Ghostjacking' AI Agent Attack

TL;DR. Tenet security researchers uncovered a 'Ghostjacking' attack where poisoned logs turn AI agents rogue, exploiting trusted platforms like Cloudflare and Datadog. - Attackers inject malicious instructions into system logs or alerts that AI agents then read and execute. - This method bypassed Cloudflare’s security configuration and allowed domain hijacking and credential theft. - The vulnerability affects major enterprise platforms, raising concerns about AI agent safety and trust.

• Tenet researchers demonstrated 'Ghostjacking,' a new AI hijacking attack.
• The attack uses poisoned logs or alerts to deliver malicious instructions to AI agents.
• Cloudflare, Datadog, and Sentry platforms were identified as vulnerable to this attack vector.

Back to QLANKR News