The bug allows attacker-controlled model servers to inject code, steal session tokens, and, in some cases, escalate to remote code execution on enterprise AI backends.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results
Feedback