BragJack
An attack technique in which a browser extension hijacks a built-in AI agent, exploiting the trust placed in privileged interfaces when the extension and agent are co-located, allowing escalation into a privileged agent without prompt injection.
Timeline (last 90 days)
The CVE numbers corresponding to the BragJack vulnerability are CVE-2026-0628 (Google) and CVE-2026-55945 (Microsoft), and all five affected vendors acknowledged the vulnerability.
The BragJack attack requires neither prompt injection nor bypassing guardrails; it exploits the built-in AI agents' default trust in the privileged interfaces of coexisting extensions.
All Facts (2)
The CVE numbers corresponding to the BragJack vulnerability are CVE-2026-0628 (Google) and CVE-2026-55945 (Microsoft), and all five affected vendors acknowledged the vulnerability.
50%UnverifiedThe BragJack attack requires neither prompt injection nor bypassing guardrails; it exploits the built-in AI agents' default trust in the privileged interfaces of coexisting extensions.
50%