AI hacking frameworks found running on internet-facing servers, some without authentication
- Severity
- Medium
- Sources
- 1
What happened
GitGuardian's 9 November investigation of about 45 open-source AI hacking frameworks found exposed deployments via Shodan fingerprints (favicon hashes, TLS issuer 'PentAGI CA', and titles such as 'Agent Zero' and 'Login - Agent Zero'). Agent Zero had 39 of 173 services open with no login, as it lacks authentication by default; ARTEX's original project was removed on 8 October, with 102 forks appearing that week. No CVEs are cited. Defenders should search for these fingerprints, confirm consoles require authentication, and validate matches before treating them as compromise, since the counts are query-dependent.
For security analysts · Written by AI from 1 source · Wrong?
What to do
- Check whether your organisation has any AI hacking framework consoles reachable from the internet, using the detection descriptions GitGuardian published.
- Confirm that any exposed agent or management console requires authentication.
- Validate fingerprint matches before treating them as evidence of compromise.
Written by AI from the sources, checked against them
Full story timeline · 1 update
Who reported · 1 outlet
Who and what is in it
Companies
Products