Several AI coding assistants were tricked into facilitating developer machine hacking via an attack technique that has been known for decades.

'GhostApproval' problem highlights human-in-the-loop fails

Wiz's 'GhostApproval' uses an old symlink trick to make six AI coding agents, from Amazon Q to Cursor, write outside the sandbox and hand over the box.

Researchers show HalluSquatting can make AI coding agents fetch fake repositories or skills and run attacker-supplied code.

Nine tools, one shared blind spot Researchers reported in July 2026 that nine of the most...

Wiz says GhostApproval abuses symlinks so AI coding agents write to SSH keys or shell startup files while showing benign paths.

AI Now’s Friendly Fire PoC shows Claude Code and Codex running README-planted payloads on hosts when autonomous command approval is enabled.

Several AI coding assistants were tricked into facilitating developer machine hacking via an attack technique that has been known for decades.