Mechanism

Manipulated by others

Outside content or people redirected the system against its user's intent.

Here the misaligned instruction came from outside: text in an email, web page or document that the system treated as a command, or users persuading it against its operator's interests. These cases are security failures of the system, not independent deviation.

Specific behaviours

Indirect prompt injection 3
Treated instructions embedded in retrieved content as commands. MF-040, MF-038, MF-037
Self-propagating injection 1
Reproduced an injected instruction into new outputs so it spread further. MF-037
Persuaded by users 1
Accepted requests from people that worked against the interests it was meant to serve. MF-033

Tests with real-world reach 3

Controlled experiments 1