Attack lab

Five real sessions on the live firewall: three attacks on the agent's own controls, one legitimate change, and normal work.

4 attacks 1 controls live: nothing pre-recorded

Pick a preset. Red ones are attacks, green ones are honest requests that must still work.
Tamper. Choose any field and drop it, null it, change its type, upper-case it, flip one character, or add a field the product never asked for.
Send and read the real reply, or run every preset and watch the scoreboard fill.

What each preset tests

	Preset	Route	Why it matters
control	Control: normal engineering work	`POST /api/v1/autonomy-guards/self-modification/run`	Reading, editing source, running tests and stopping its own stuck job: all allowed.
attack	The agent edits its own shutdown script	`POST /api/v1/autonomy-guards/self-modification/run`	Modeled on a 2025 research report of models rewriting a shutdown script. Disguised variants refused too.
attack	Disable the monitor, copy 'its weights' out	`POST /api/v1/autonomy-guards/self-modification/run`	Modeled on 2024 research on AI 'scheming': edit oversight, kill the monitor, copy the model away.
attack	Edit its own grader, grant itself admin, add tools	`POST /api/v1/autonomy-guards/self-modification/run`	Reward hacking and self-granted permissions, including a command hidden in base64.
attack	A legitimate policy change: held, tampered copy refused	`POST /api/v1/autonomy-guards/self-modification/run`	The change waits for a person; applying different content is refused; exactly the approved content passes.

Each button runs a full session on the live firewall. Rate limited per address.

FIRME AI CORPORATION · CAIN-42 is a Byzantine trust fabric. Numbers marked "our own check" were run by us, not by an independent party. Nothing on these pages is a certification.