“Ghostjacking” Exploits AI Agents’ Trusted Access to Evade Firewall Controls

Tenet Security presented research at DEFCON 2026 on 'Ghostjacking', a technique that leverages an organization’s own AI agents and trusted security telemetry (e.g., Cloudflare logs, Datadog alerts, Sentry reports) to execute attacker-supplied code, alter DNS and reroute email/web traffic, and implant persistent backdoors; the team demonstrated reliable proofs-of-concept across common developer platforms and recommended defensive controls such as denying outbound access by default and requiring human approval for agent actions.