GenAI Security Project
Comment: In accordance with Wikipedia's Conflict of interest guideline, I disclose that I have a conflict of interest regarding the subject of this article. Owaspsclinton (talk) 01:40, 11 October 2026 (UTC)
| File:OWASP GenAI Security Project Logo.png | |
| Formation | May 2023 (as the OWASP Top 10 for LLM Applications project) |
|---|---|
| Founders | Scott Clinton, Steve Wilson |
| Type | Open-source community project |
Members | 30,000+ |
Board co-chairs | Scott Clinton, Steve Wilson |
Parent organization | OWASP Foundation |
| Website | genai |
GenAI Security Project
Global open-source community for generative and agentic AI security
The OWASP GenAI Security Project is a leading global open-source community of security practitioners, CISOs, and industry experts affiliated with the Open Worldwide Application Security Project (OWASP) that identifies and documents security and safety risks mitigations and solutions in generative artificial intelligence, including large language models (LLMs), agentic AI systems and AI-driven applications, and publishes guidance for mitigating them.[1] It grew out of the OWASP Top 10 for LLM Applications, a risk list started by Steve Wilson in May 2023, and was co-founded in its present form by Scott Clinton and Wilson. It took its current name in March 2025, when OWASP designated it a flagship project.[2][3][4]
The project's publications are written by security practitioner, experts and CISO volunteers and released under open licences. They include risk lists, threat taxonomies, governance checklists, red teaming guidance and open-source tools, and are aimed at developers, security practitioners, chief information security officers and policymakers.[1][5] As of 2026 its work is organised into eight initiatives,[6] and the project reports more than 30,000 community members in over 18 countries and more than 45 publications.[7]
History
Top 10 for LLM Applications (2023 to 2024)
The project originated in May 2023, when Steve Wilson founded the OWASP Top 10 for LLM Applications and drafted its first version, working with a small group of security practitioners and AI researchers. That document laid the foundation for the later GenAI Security Project. It followed the rapid adoption of LLM-based products and the absence of structured security guidance for them, and used the format of OWASP's existing Top 10 risk lists.[2] Version 1.1 of the list followed later in 2023.[8]
By early 2024 the project reported more than 600 contributing experts from over 18 countries and close to 8,000 community members.[2] During 2024 it broadened from the risk list into governance and operational guidance, publishing among others:
- the LLM Cybersecurity and Governance Checklist for security leaders (April 2024);
- the Guide for Preparing and Responding to Deepfake Events, the first output of its threat intelligence work (September 2024);
- the AI Security Center of Excellence Guide and the first LLM and Generative AI Security Solutions Landscape (October 2024);
- the 2025 edition of the Top 10 for LLM Applications, released in November 2024 together with a sponsorship programme.[2]
The Agentic Security Initiative, concerned with autonomous AI agents, was launched in December 2024, and a GenAI Red Teaming Guide was published in January 2025.[2]
GenAI Security Project (2025 to present)
On March 27, 2025, OWASP announced that the Top 10 for LLM project had been renamed the OWASP GenAI Security Project and promoted to flagship status. The Top 10 list became one initiative within the larger project.[3] Scott Clinton and Wilson are the co-founders of the GenAI Security Project and co-chair its board; Clinton is credited by the project with leading its strategy, operations and growth.[4]
In December 2025 the project published the OWASP Top 10 for Agentic Applications, a list of risks specific to autonomous agents, described by the project as developed with more than 100 contributors.[9]
In March 2026, ahead of the RSA Conference, the project reported more than 25,000 members and announced further publications, including GenAI Data Security Risks and Mitigations for 2026, a Guide for Secure MCP Server Development, vendor evaluation criteria for AI red teaming and an open-source AI bill of materials generator.[10]
In September 2026 the project announced the 2026 edition of the Top 10 for LLM Applications, which it said was the first edition to test contributor rankings against a large set of recorded incidents. At the same time it announced that the Agent Control Standard (ACS), a specification for runtime control of AI agents, had been donated to the project, and published a crosswalk mapping its guidance to other security and compliance frameworks. The project reported more than 30,000 members on LinkedIn at that time.[11]
Mission
According to its charter, the project aims to give organisations, security professionals, AI practitioners and policymakers practical guidance and tools for the secure development, deployment and governance of generative AI systems. The charter lists five objectives:[1]
- identifying and documenting risks, including maintaining the Top 10 for LLM Applications;
- developing security best practices and mitigations;
- applied research and collaboration with academic, industry and government partners;
- education, training and translation of its publications;
- supporting enterprise adoption and AI security governance.
Initiatives
Work is carried out by initiatives, which are working groups approved by the project's core team and board on the basis of a written proposal.[5] As of 2026 there are eight:[6][4]
| Initiative | Focus | Leads |
|---|---|---|
| Top 10 for LLM and GenAI | The most critical security risks in LLM and generative AI applications | Rock Lambros, Steve Wilson |
| Agentic Security | Security of autonomous agents and multi-step AI workflows; publishes the Top 10 for Agentic Applications | Ron F. Del Rosario, John Sotiropoulos |
| AI Bill of Materials (AIBOM) | Standards and tooling for AI supply chain transparency | Helen Oakley, Dmitry Raidman |
| Data Security | Protection of training and retrieval data from leakage and tampering | Emmanuel Guilherme, Scott Clinton |
| AI Security Governance | Guidance for adopting and governing generative AI securely | Sandy Dunn |
| AI Security Solutions | A regularly updated landscape and directory of AI security tools | Scott Clinton |
| AI Threat Intelligence and Response | Tracking misuse of generative AI by attackers; incident response guidance | Rachel James, Bryan Nakayama |
| AI Red Teaming | Methods for adversarial testing and evaluation of generative AI systems | Sonu Kumar, Jason Ross |
Publications
Top 10 for LLM Applications
The Top 10 for LLM Applications is the project's original and best-known publication. Editions have been issued for 2023 to 2024, 2025 and 2026.[8][2][11] The 2026 edition includes mappings to NIST guidance, MITRE ATLAS, the Common Weakness Enumeration and the project's own agentic list.[11]
Top 10 for Agentic Applications
Published in December 2025, this list covers ten risk categories, numbered ASI01 to ASI10, for systems in which AI agents plan, retain memory, call tools and act with delegated authority.[9] It was released alongside related work from the Agentic Security Initiative: Agentic AI Threats and Mitigations, A Practical Guide to Securing Agentic Applications, The State of Agentic Security and Governance, an Agentic Security Solutions Landscape and FinBot, a capture-the-flag reference application for practising agentic security.[9]
Other resources
Other publications include the LLM Cybersecurity and Governance Checklist, the AI Security Center of Excellence Guide, the GenAI Red Teaming Guide, the Guide for Preparing and Responding to Deepfake Events, periodic round-ups of AI-related security incidents, the AI Security Solutions Landscape, the AI Security Solutions Directory, the Agent Control Standard and the GenAI Security Industry Framework Crosswalk.[2][11] Publications announced in 2026 also include GenAI Data Security Risks and Mitigations for 2026, the Guide for Secure MCP Server Development, Vendor Evaluation Criteria for AI Red Teaming and the OWASP AIBOM Generator, an open-source tool for producing AI bills of materials.[10] Project documents have been translated into several languages, including Portuguese, German, French, Hindi, Persian, Simplified Chinese, Japanese and Spanish.[2]
Documents are released under the Creative Commons Attribution-ShareAlike 4.0 licence; code is released under permissive licences such as MIT, BSD or Apache 2.0.[5]
Use in government guidance
Publications from the project have been cited in security guidance issued by government agencies and standards bodies in several countries. These are mostly voluntary guidance documents and technical reports, not binding regulation. Examples include:
| Jurisdiction | Document | Issuer | Date | Project material cited |
|---|---|---|---|---|
| Canada | Generative and Agentic Artificial Intelligence: Implications for Technology, Cyber Security, and Operational Resilience (Technology Risk Bulletin)[12] | Office of the Superintendent of Financial Institutions | July 2026 | Lists the Top 10 for LLM Applications (2025) and State of Agentic AI Security and Governance among its related links |
| Canada | Securely deploying AI at the network edge (ITSP.80.101)[13] | Canadian Centre for Cyber Security | July 2026 | Links the Top 10 for Agentic Applications 2026 as further reading |
| Europe | ETSI TR 104 128: Guide to Cyber Security for AI Models and Systems[14] | ETSI Technical Committee on Securing AI | May 2025 | Cites the Top 10 for LLM Applications, the LLM Cybersecurity and Governance Checklist, the AI Security Center of Excellence Guide, the Solutions Landscape and the deepfake guide |
| Germany | Generative AI Models: Opportunities and Risks for Industry and Authorities (version 2.0)[15] | Federal Office for Information Security (BSI) | January 2025 | Cites the 2023 Top 10 for LLM Applications |
| Germany and France | Design Principles for LLM-based Systems with Zero Trust[16] | BSI and ANSSI | August 2025 | Cites the Top 10 for LLM Applications 2025 and the GenAI Red Teaming Guide in its bibliography |
| Japan | Guide to Evaluation Perspectives on AI Safety (version 1.10)[17] | Japan AI Safety Institute | March 2025 | Cites the Top 10 for LLM Applications as a source in its security section |
| Japan | Guide to Red Teaming Methodology on AI Safety (version 1.10)[18] | Japan AI Safety Institute | March 2025 | Cites the Top 10 for LLM Applications in its attack-scenario step |
| Singapore | Guidelines and Companion Guide on Securing AI Systems[19] | Cyber Security Agency of Singapore | October 2024 | The Companion Guide cites the Top 10 for LLM Applications against three controls |
| Singapore | Model AI Governance Framework for Agentic AI (version 1.5)[20] | Infocomm Media Development Authority | May 2026 | Cites Agentic AI: Threats and Mitigations |
| Singapore | Securing Agentic AI: An Addendum to the Guidelines and Companion Guide on Securing AI Systems[21] | Cyber Security Agency of Singapore | June 2026 | Uses the project's agentic threat taxonomy and cites the Top 10 for Agentic Applications 2026 |
| United Kingdom | Implementation Guide for the AI Cyber Security Code of Practice[22] | Department for Science, Innovation and Technology | January 2025 | Cites the Top 10 for LLM Applications throughout, with the governance checklist, Solutions Landscape, deepfake guide and Center of Excellence guide |
| United States, United Kingdom, Canada, Australia and New Zealand | Careful Adoption of Agentic AI Services[23] | Canadian Centre for Cyber Security with NCSC, CISA, NSA, the Australian Signals Directorate and NCSC-NZ | April 2026 | Names the project as a threat-modelling taxonomy and lists both Top 10s as resources |
The UK Implementation Guide was written for the department by John Sotiropoulos, a member of the project's board, and the ETSI technical report is its standardised form.[4][14]
Events and education
The project has held a half-day AI security summit at the RSA Conference in San Francisco in three consecutive years: 2024, 2025 and 2026.[24][25][10] The 2026 edition was accompanied by an open workshop and a hackathon built on the FinBot capture-the-flag application.[10]
It also runs virtual summits on generative AI application security and risk, the most recent on May 27, 2026.[26] Other events have included an agentic AI summit in London in December 2025, at which the Top 10 for Agentic Applications was launched, a summit at Infosecurity Europe in June 2026, and a summit and hackathon in Dubai.[26][27] The project publishes recordings of its open community meetings and other learning videos, and a newsletter.[7]
Organisation
The project follows a meritocratic governance model. A board of directors is its highest decision-making body, and a Core Project Management Team, made up of initiative leads and operational or technical leads, handles day-to-day coordination. Routine decisions are made by lazy consensus, with a 72-hour objection period; strategic matters go to a vote of initiative leads, the core team or the board. Participation is open to volunteers.[5]
As of 2026 the board of directors consists of:[4]
- Scott Clinton: co-chair and co-founder
- Ads Dawson: technical lead
- Sandy Dunn: co-lead, AI Security Governance initiative
- John Sotiropoulos: co-lead, Agentic Security Initiative
- Steve Wilson: co-chair and co-founder; founder of the Top 10 for LLM Applications
The project's initiative leads and other core team members are listed in full on the leadership page of its website.[4]
Expert review boards
Under the governance model, an initiative lead may form an expert review board: a group of recognised specialists, who need not be regular contributors, that reviews and helps validate the initiative's output before publication.[5] The board that reviewed the Top 10 for Agentic Applications included reviewers from the Alan Turing Institute, Amazon Web Services, Cisco, the Institute for Ethical AI, Microsoft, the National Institute of Standards and Technology (NIST), Oracle and Zenity.[28] The project's announcement of the list also named the European Commission among the bodies represented.[9]
Agentic Research Council
The Agentic Research Council (ARC) was launched by the project's Agentic Security Initiative on June 4, 2026, at the OWASP GenAI Security Summit held during Infosecurity Europe in London. It is chaired by John Sotiropoulos.[29][30] The council is intended to bring together researchers from academia, industry and government so that research on agentic AI security reaches practitioners faster than conventional standards cycles allow. Its stated plans include maintaining a public pipeline of research topics, running working groups, sponsoring doctoral research and feeding results into the project's guidance and tools. Its first research topic is the security of multi-agent systems.[29][27]
Sponsors and partners
The project is funded in part by corporate sponsors under a programme introduced in November 2024.[2] Sponsors announced in March 2026 were Apiiro, Capsule, F5, Fujitsu, NeuralTrust, Starseer, Straiker and Tellus Digital.[10] In September 2026 the project named F5 and WitnessAI as new gold sponsors and Evoke Security and Mondoo as new silver sponsors.[11] A current list is kept on the project's website.[31]
The project has worked with bodies including MITRE, NIST and UK government departments.[2]
See also
References
- ↑ 1.0 1.1 1.2 "OWASP Gen AI Security Project Charter". OWASP GenAI Security Project. Retrieved 2026-10-10.
- ↑ 2.00 2.01 2.02 2.03 2.04 2.05 2.06 2.07 2.08 2.09 "OWASP Gen AI Security Project Introduction and Background". OWASP GenAI Security Project. Retrieved 2026-10-10.
- ↑ 3.0 3.1 "Project OWASP Promotes GenAI Security Project to Flagship Status" (Press release). OWASP Foundation. 2025-03-27. Retrieved 2026-10-10 – via PR Newswire.
- ↑ 4.0 4.1 4.2 4.3 4.4 4.5 "Leadership & Contributors". OWASP GenAI Security Project. Retrieved 2026-10-10.
- ↑ 5.0 5.1 5.2 5.3 5.4 "Project Governance". OWASP GenAI Security Project. Retrieved 2026-10-10.
- ↑ 6.0 6.1 "Initiatives". OWASP GenAI Security Project. Retrieved 2026-10-10.
- ↑ 7.0 7.1 "OWASP GenAI Security Project". OWASP GenAI Security Project. Retrieved 2026-10-10.
- ↑ 8.0 8.1 "OWASP Top 10 for LLM Applications 2023/24". OWASP GenAI Security Project. Retrieved 2026-10-10.
- ↑ 9.0 9.1 9.2 9.3 "OWASP GenAI Security Project Releases Top 10 Risks and Mitigations for Agentic AI Security" (Press release). OWASP GenAI Security Project. 2025-12-10. Retrieved 2026-10-10 – via PR Newswire.
- ↑ 10.0 10.1 10.2 10.3 10.4 "OWASP GenAI Security Project Expands AI Security Frameworks Ahead of RSA 2026, Celebrates Continued Sponsor Support" (Press release). OWASP GenAI Security Project. 2026-03-19. Retrieved 2026-10-10 – via PR Newswire.
- ↑ 11.0 11.1 11.2 11.3 11.4 "OWASP GenAI Security Project Releases 2026 Top 10 for LLM Applications, Debuts Agent Control Standard and New Resources for Securing Generative and Agentic AI" (Press release). OWASP GenAI Security Project. 2026-09-02. Retrieved 2026-10-10 – via PR Newswire.
- ↑ "Generative and Agentic Artificial Intelligence: Implications for Technology, Cyber Security, and Operational Resilience (Technology Risk Bulletin)". Office of the Superintendent of Financial Institutions. 2026-07. Retrieved 2026-10-10. Check date values in:
|date=(help) - ↑ "Securely deploying AI at the network edge (ITSP.80.101)". Canadian Centre for Cyber Security. 2026-07. Retrieved 2026-10-10. Check date values in:
|date=(help) - ↑ 14.0 14.1 "ETSI TR 104 128: Guide to Cyber Security for AI Models and Systems" (PDF). ETSI Technical Committee on Securing AI. 2025-05. Retrieved 2026-10-10. Check date values in:
|date=(help) - ↑ "Generative AI Models: Opportunities and Risks for Industry and Authorities (version 2.0)" (PDF). Federal Office for Information Security (BSI). 2025-01. Retrieved 2026-10-10. Check date values in:
|date=(help) - ↑ "Design Principles for LLM-based Systems with Zero Trust". BSI and ANSSI. 2025-08-11. Retrieved 2026-10-10.
- ↑ "Guide to Evaluation Perspectives on AI Safety (version 1.10)" (PDF). Japan AI Safety Institute. 2025-03-28. Retrieved 2026-10-10.
- ↑ "Guide to Red Teaming Methodology on AI Safety (version 1.10)" (PDF). Japan AI Safety Institute. 2025-03-31. Retrieved 2026-10-10.
- ↑ "Guidelines and Companion Guide on Securing AI Systems". Cyber Security Agency of Singapore. 2024-10-15. Retrieved 2026-10-10.
- ↑ "Model AI Governance Framework for Agentic AI (version 1.5)" (PDF). Infocomm Media Development Authority. 2026-05-20. Retrieved 2026-10-10.
- ↑ "Securing Agentic AI: An Addendum to the Guidelines and Companion Guide on Securing AI Systems". Cyber Security Agency of Singapore. 2026-06-17. Retrieved 2026-10-10.
- ↑ "Implementation Guide for the AI Cyber Security Code of Practice" (PDF). Department for Science, Innovation and Technology. 2025-01-31. Retrieved 2026-10-10.
- ↑ "Careful Adoption of Agentic AI Services" (PDF). Canadian Centre for Cyber Security with NCSC, CISA, NSA, the Australian Signals Directorate and NCSC-NZ. 2026-04-30. Retrieved 2026-10-10.
- ↑ "OWASP AI Security Summit: Safeguarding AI with OWASP Top 10 for LLMs & Gen AI". RSA Conference. 2024. Retrieved 2026-10-10.
- ↑ "OWASP Gen AI Security Project Announces Nine New Sponsors and Major RSA Conference Presence to Advance Generative AI Security" (Press release). OWASP GenAI Security Project. 2025-04-17. Retrieved 2026-10-10 – via PR Newswire.
- ↑ 26.0 26.1 "Events". OWASP GenAI Security Project. Retrieved 2026-10-10.
- ↑ 27.0 27.1 "Top 3 Highlights from the OWASP GenAI Security Summit at Infosecurity Europe". Infosecurity Europe. 2026-06-19. Retrieved 2026-10-10.
- ↑ "OWASP Top 10 for Agentic Applications: The Benchmark for Agentic Security in the Age of Autonomous AI". OWASP GenAI Security Project. 2025-12-09. Retrieved 2026-10-10.
- ↑ 29.0 29.1 Poireault, Kevin (2026-06-01). "Infosecurity Europe: OWASP Forms New Agentic Research Council". Infosecurity Magazine. Retrieved 2026-10-10.
- ↑ Sotiropoulos, John (2026-06). "Delighted to officially launch and chair the Agentic Research Council". LinkedIn. Retrieved 2026-10-10. Check date values in:
|date=(help) - ↑ "Sponsors and Supporters". OWASP GenAI Security Project. Retrieved 2026-10-10.
External links
Category:OWASP Category:Computer security organizations Category:Artificial intelligence safety Category:Organizations established in 2023
This article "GenAI Security Project" is from Wikipedia. The list of its authors can be seen in its historical and/or the page Edithistory:GenAI Security Project. Articles copied from Draft Namespace on Wikipedia could be seen on the Draft Namespace of Wikipedia and not main one.
