Four vulnerabilities in CrewAI could be chained together via prompt injection for sandbox escape, remote code execution, and ...
Claude extension flaw enabled silent prompt injection via XSS and weak allowlist, risking data theft and impersonation until ...
Command injection in Codex and a hidden outbound channel in ChatGPT exposed risks of credential theft and covert data ...
Authentication Failures (A07) show the largest gap in the dataset: a 48-percentage-point difference between leaders and the field. Leaders fix at nearly 60%, while the field sits at roughly 12%.
The path traversal flaw, allowing access to arbitrary files, adds to a growing set of input validation issues in AI pipelines.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results