The operational threat-intelligence layer of the OWASP GenAI Security Project.
Established in 2024, the Threat Intelligence Initiative was founded and is co-led by Rachel James (GitHub · LinkedIn) and Bryan Nakayama (LinkedIn). Learn more on the OWASP GenAI initiative page.
🚧 This repository is in a building phase. The initiative itself is well established — what's new is the AI-CTI platform hosted here: the collection pipeline, public catalog, website, and STIX/TAXII feed described below are in active development. Expect this repo to change rapidly. Contributions and reviewers welcome.
The rapid adoption of LLMs and Generative AI has opened a threat landscape that traditional CTI approaches don't fully cover. This initiative fills the tactical middle layer between high-level risk frameworks (the OWASP Top 10 for LLM Applications, the CISO Governance Checklist) and the hands-on intelligence that security operations teams need to detect and respond to real-world attacks.
We produce threat-informed, actionable guidance for the practitioners defending GenAI systems in production — SOC analysts, CTI analysts, threat hunters, incident responders, detection engineers, security researchers, and MLSecOps teams.
- Category A — Threat actors using AI to conduct attacks. LLM-enhanced phishing, malware development, reconnaissance, and deepfake-enabled fraud. Mapped to MITRE ATT&CK plus a custom LLM-TTP taxonomy.
- Category B — Threat actors attacking AI systems. Confirmed adversary (not researcher) activity against AI/ML/LLM systems — prompt injection, RAG poisoning, model supply-chain compromise. Mapped to MITRE ATLAS.
The golden rule: only confirmed, attributed, cited activity is published. Nothing goes live until a human analyst marks it Analyst Confirmed.
A sustainable, OWASP-owned AI-CTI platform that turns curated research into a standards-based intelligence service:
- AI-assisted collection pipeline — automated discovery of new reports (with sources) and structured extraction into candidate records: actor, aliases, brief, TTPs, and citations.
- Human review & approval — a lightweight dashboard with notifications, where analysts confirm or correct each draft and set the publish gate.
- Public website — published, confirmed entries for both categories.
- STIX 2.1 / TAXII 2.1 feed — a machine-readable feed other organizations can ingest into their threat-intelligence platforms (OpenCTI, MISP, SIEMs).
Data is Git-centric: approved records live as structured YAML/JSON in this OWASP org, and the website, STIX bundles, and TAXII feed all generate from them. Every automated draft passes through human analyst review before publication.
- Threat intelligence briefings on active campaigns, emerging attack patterns, and adversary TTPs
- IOCs / IOAs for GenAI-specific threats (prompt-injection patterns, jailbreak attempts, model extraction and data-exfiltration signatures)
- Detection-engineering guidance (SIEM rules, detection logic, monitoring frameworks)
- Incident-response playbooks tailored to GenAI incidents
- Threat-actor tracking of AI weaponization and associated TTPs
- Real-world attack case studies and lessons learned
STIX / TAXII · MITRE ATT&CK · MITRE ATLAS · industry-standard IOC formats. We reference authoritative sources rather than duplicating them, and maintain vendor/product neutrality.
| Phase | What | Status |
|---|---|---|
| 0 | Foundations: OWASP repo, schema, migrate existing catalog | 🔨 In progress |
| 1 | Publish existing corpus: website + STIX/TAXII feed | ⬜ Planned |
| 2 | AI collection pipeline + spend caps | ⬜ Planned |
| 3 | Review dashboard + notifications | ⬜ Planned |
| 4 | Agentic-AI incidents data exchange | ⬜ Planned |
| 5 | Hardening, backups & maintainer handoff | ⬜ Planned |
The initiative organizes its work into groups spanning threat research, detection engineering, incident response, and adversary tracking, with regular community meetings to coordinate research and discuss emerging threats.
- Reviewers — ~1 hr/week confirming entries. No coding required.
- STIX / CTI builders — help with the TAXII feed and STIX mapping.
- Pipeline / dashboard builders — building the collection pipeline and review dashboard.
Contribution guidelines and an open-source license for this repository are being finalized as part of the building phase.
OWASP GenAI Security Project — Threat Intelligence Initiative.
Github Location for the GenAI Security Project's Threat Intelligence Initiative