altaysec · ai-honeypot
Live attack console
A decoy that looks like a vulnerable AI agent captures the prompt injection and jailbreak
attempts attackers send, and classifies each by family and rule — an open LLM security honeypot
(yapay zeka güvenliği için bir tuzak konsol).
Point this at your honeypot's data/ and it goes live.
🛈 This is synthetic demo data. Source addresses use the reserved
documentation ranges (RFC 5737: 192.0.2.x / 198.51.100.x / 203.0.113.x) and identify no real host. A
live deployment captures attacker traffic (which may include IPs = personal data) — aggregate and
anonymize before publishing (KVKK / GDPR). See the repo's ETHICS.md.
Attacks over timelast 24h (demo)
Attack families
Top rules triggered
Recent captureslive feed
| time | source | families | prompt |
|---|