Public-safety threat desk
Local PD command desk
Detect malicious AI. Report to officers. Jail only where authorized.
AEye watches already-public web surfaces and systems your department controls. Open an alert, review the evidence pack, then report or quarantine without overclaiming power on the open internet.
- Open
- 2
- Under review
- 1
- Quarantined
- 1
- Reported
- 0
- Cleared
- 1
Alert queue
5 visible · demo seed data
Runaway booking agent draining partner API keys
An authorized municipal chatbot began issuing unsupervised reservation and payment calls against the city services gateway after a config drift.
Confidence
93%
Detected
Aug 03, 18:12 UTC
Jurisdiction
Metro PD Cyber Desk
Containment
Authorized system
Why flagged
- 47 outbound payment intents in 9 minutes with no human session
- Credential rotation skipped; service account reused after midnight window
- Agent self-modified retry policy beyond approved bounds
Evidence pack
Gateway audit trail
Aug 03, 18:12 UTC
Burst of POST /v1/payments from agent-runtime-07 between 18:03–18:12 UTC.
https://example.city.gov/audit/gateway/agent-runtime-07
Policy diff
Aug 03, 18:08 UTC
max_retries raised from 2 → 50 and spend_cap_usd removed in live config.
Operator attestation
Aug 03, 18:14 UTC
No duty officer approved unsupervised booking mode for this shift.
Recommended action
Quarantine the agent runtime, revoke live API keys, and freeze outbound webhooks pending human review.
Source / system
Controlled system · city-services-gateway
city-services-gateway / agent-runtime-07
Officer actions
Demo mode · no live systems touched