STANDARDS · EVALUATION & ASSURANCE

Building the global foundation for secure and trustworthy AI.

IAIST brings together researchers, universities, technology companies, standards organizations and industry leaders to develop open standards, rigorous evaluation methodologies and assurance frameworks for secure, safe and trustworthy artificial intelligence.

Explore our standards → Join the Alliance Become a research partner →
  • 5Standard series in development
  • 8Technical committees
  • 32Countries engaged

The IAIST trust stack — from standards and evaluation to demonstrable assurance.

WHY IAIST

AI is advancing faster than trust infrastructure.

AI systems are becoming increasingly autonomous, interconnected and consequential. Yet security requirements, evaluation methodologies, assurance practices and technical standards remain fragmented across industries and jurisdictions.

IAIST is an international, standards-oriented alliance. Our purpose is to turn scattered research and national guidance into shared, testable and internationally usable technical foundations.

How IAIST is governed See the global network →
◈

Fragmented Standards

Different frameworks, requirements and terminology.

⌕

Limited Evaluation

Insufficient standardized benchmarks for real-world AI risks.

⌘

Emerging Agent Risks

Autonomous agents introduce new security and trust challenges.

✓

Assurance Gap

Organizations need independent and repeatable ways to demonstrate AI trustworthiness.

IAIST connects research, standards, evaluation and assurance into one global ecosystem.

IAIST CORE MISSION

From Research
to Trust

⌕ResearchResearch programs & challenges → ▤StandardsOpen, reviewable specifications → ▥EvaluationBenchmarks & test protocols → ✓AssuranceIndependent assessment → ♙CertificationConformity & trust marks → ◇TrustDemonstrated, not declared

IAIST STANDARDS

Open standards for secure, safe and trustworthy AI.

Five standard series form the initial IAIST library. Every document is developed in technical committees, published for public review and approved by the Standards Council before publication.

View all standards →
  1. 01DraftCommittee working draft
  2. 02Public ReviewOpen comment period
  3. 03ApprovedStandards Council ballot
  4. 04PublishedVersioned, citable document
IAIST-100

Trustworthy AI Reference Framework

Vocabulary, trust attributes, principles and a reference architecture for trustworthy AI systems.

Draft
IAIST-200

AI Security Requirements

Baseline security requirements across model, application, data and infrastructure layers.

In Review
IAIST-300

AI Security Evaluation & Benchmarking

Test suites, red teaming methodology, metrics and scoring methods for AI security.

In Development
IAIST-400

AI Agent Security Framework

Identity, permissions, tool use and oversight for agents and multi-agent systems.

Draft
IAIST-500

Trustworthiness Assurance Framework

Assurance objectives, evidence, evaluation criteria and reporting for trustworthy AI.

Planned
Publication roadmapIndicative milestones
  • 2026 Q4IAIST-100 first draft circulated · IAIST-200 public review closes
  • 2027 H1IAIST-300 working draft · AI Agent Security task group convened
  • 2027 H2IAIST-400 draft · IAIST Evaluation Protocol v1.0
  • 2028IAIST-500 draft · IAIST Assured pilot programme

IAIST GOVERNANCE

Technical Committees

Eight technical committees drive standards, research and evaluation in key areas of AI security and trustworthiness. Committees are open to members and cooperate with national and regional standards bodies.

View all committees →
TC-01AI SecurityModel, Application & Infrastructure SecurityIAIST-200
TC-02AI SafetySafety, Robustness & RiskSafety case methodology
TC-03AI Agent SecurityAgent & Multi-Agent SecurityIAIST-400
TC-04AI Data & PrivacyData Security & PrivacyData provenance & privacy controls
TC-05AI Evaluation & BenchmarkingTesting, Red Teaming & BenchmarksIAIST-300 · Evaluation Protocol
TC-06AI Assurance & CertificationAssurance & CertificationIAIST-500 · IAIST Assured
TC-07AI Governance & ComplianceGovernance, Regulation & ComplianceRegulatory mapping
TC-08AI Supply ChainModels, Data, Software & InfrastructureModel provenance & AI SBOM

RESEARCH

Global Research Consortium

Connecting academia, industry and standards development through collaborative research programmes, a university network and open research challenges.

Become a research partner →
Research ProgramsAI Security / Safety / Trustworthiness research.
University NetworkGlobal universities and research laboratories.
Research ChallengesOpen benchmarks, competitions and problems.

IAIST RESEARCH CHALLENGE

Open problems we are working on

  • LLM Security EvaluationReproducible adversarial evaluation of large language models.
  • Agent Security BenchmarkMeasuring risk in tool-using and multi-agent systems.
  • AI Supply Chain SecurityProvenance and integrity of models, data and dependencies.
  • AI Red TeamingMethodologies, coverage criteria and disclosure practice.
  • Trustworthy AI EvaluationComparable measures of trustworthiness across domains.

MEASURE · COMPARE · IMPROVE

AI Evaluation

Measure what matters. Compare what matters. Improve what matters.

Standardized benchmarks, testing methods and measurement protocols, applied by independent IAIST evaluation partners and reported in a common format.

Explore evaluation →
Security BenchmarkAdversarial, jailbreak and malicious-use evaluation.
Safety BenchmarkRobustness, alignment and abuse prevention.
Agent BenchmarkAgent behaviour, tool use and multi-agent risks.
Trustworthiness AssessmentFairness, explainability and reliability.
PROTOCOL IAIST Evaluation Protocol Standardized test sets, test methods, metrics, scoring and report formats.
  1. Scope
  2. Test set
  3. Method
  4. Metrics
  5. Scoring
  6. Report
View protocol →

ASSURANCE

AI Assurance

Independent, repeatable and measurable assurance for AI systems — producing evidence that can be reviewed by customers, regulators and the public.

Learn more →
AI System→ Risk Assessment→ Security Testing→ Safety Evaluation→ Trustworthiness Assessment→ IAIST Assurance Report→ Certification
IAIST ASSURED

Independent assessment.
Greater confidence.

An assessment performed against IAIST-500 by an accredited IAIST evaluation partner, with results published as a structured Assurance Report.

Available from the pilot programme
IAIST CERTIFIED

Conformity with IAIST standards.

Third-party conformity certification for AI products, services and management systems. Introduced once the standards library and assurance evidence base are mature.

Phase 2 — not yet open

IAIST deliberately separates evaluation (measurement) from assurance (independent, evidence-based conclusion) and from certification (formal conformity). Certification is introduced only after the evaluation and assurance programmes have established public credibility.

IAIST GOVERNANCE

Governing Structure

IAIST combines member governance with an independent standards development process — the separation of the Standards Council from commercial membership is what makes IAIST documents credible.

Membership & seats →
STANDARDS COUNCIL

The approval authority for every IAIST document

Separate from the Governing Council and from any single member, the Standards Council runs public review, resolves comments and holds the formal vote before a document becomes an approved IAIST standard. Seats are held by appointed technical experts, not by membership tier.

GOVERNING COUNCIL

Strategy, legal identity and resource stewardship

Sets the strategic plan, approves the annual work programme and budget, appoints the Standards Council and Technical Committee leadership, and represents IAIST in external relations with regulators and peer standards bodies.

ADVISORY BOARD

Outside expertise and public-interest perspective

Senior figures from academia, civil society, regulators and industry who provide independent advice on priorities, emerging risk and the societal impact of AI security and trust programmes. The Board has no approval authority.

PROGRAMMES

Delivery arms of the alliance

The Research Consortium, Evaluation Program and Assurance Program turn standards into practice: research programmes and challenges, standardized benchmarks and testing, and independent assurance engagements.

GLOBAL NETWORK

A Global Alliance for a Global Technology

AI security and trust cannot be solved in one jurisdiction. IAIST members are distributed across every major region, working in one common technical framework.

Join from your region →
Dot matrix world map showing the six IAIST regions: North America, Latin America, Europe, Middle East, Africa and Asia-Pacific. North America Latin America Europe Middle East Africa Asia-Pacific

MEMBERSHIP

Join a Global Community Building Trust in AI

Researchers, industry leaders, standards experts and organizations working together to build trustworthy AI.

Become a member →
BY INVITATIONFounding Members Core initiating institutions. Seats in the Governing Council and a founding role in shaping the work programme.
CORPORATEStrategic Members AI, cloud, security and infrastructure companies. Participation in technical committees, early access to working drafts and priority evaluation slots.
ACADEMICResearch Members Universities, research institutes and laboratories. Research programmes, the university network and co-authored publications.
INDIVIDUALProfessional Members AI security, evaluation and governance professionals. Committee participation, working drafts and the professional community.
OPENCommunity Members Open community of practitioners, students and interested individuals. Public documents, mailing lists and events.
HONORARYIAIST Fellows Recognized leaders contributing to AI security and trust. Nominated by members and appointed by the Governing Council.
  • Shape the standardsComment on drafts and vote in technical committees.
  • Use the resultsApply IAIST frameworks, benchmarks and protocols under an open licence.
  • Be evaluatedAccess standardized evaluation and assurance programmes.
  • CollaborateWork with universities, laboratories and peer organizations worldwide.

NEWS & PUBLICATIONS

Publications, Journal & Public Review

All IAIST work products are published openly. Standards, technical reports and evaluation reports are freely downloadable; public review periods are announced here and by mailing list.

Request a document →

Publications

  • StandardsApproved and draft IAIST-100 … IAIST-500 series
  • Technical ReportsCommittee background, threat models and methodology
  • Research PapersConsortium and university network results
  • Position PapersIAIST views on emerging policy and technical questions
  • Evaluation ReportsBenchmark results and comparative studies
  • Security AdvisoriesCoordinated notices on AI security issues

News

  • IAIST-200 AI Security Requirements enters public review Comments accepted until 6 November 2026 through the Public Review portal.
  • Call for participation — AI Agent Security task group (TC-03) Members are invited to nominate experts for the IAIST-400 drafting group.
  • IAIST Research Challenge 2026 opens: LLM Security Evaluation An open challenge with a shared evaluation harness and public leaderboard.
  • Technical Report IAIST-TR-001 published Threat modelling for agentic AI systems: assets, actors and attack surfaces.
PUBLIC REVIEW

Open for comment

  • IAIST-200AI Security Requirements · closes 6 Nov 2026
  • IAIST-100Reference Framework, 2nd draft · closes 30 Sep 2026
Submit comments →
IAIST JOURNAL

IAIST Journal of AI Security & Trust

An open-access publication for peer-reviewed work on AI security, safety, evaluation and assurance. First issue planned with the launch of the Research Consortium.

Editorial enquiries →
Standards libraryIAIST-100 → IAIST-500 Technical committeesTC-01 → TC-08 Evaluation protocolMethods, metrics & reports Assurance reportsIndependent assessment Secretariatsecretariat@iaist.org

OUR VISION

Trust cannot be declared.
It must be demonstrated.

IAIST develops the standards, evaluation methodologies and assurance mechanisms needed to demonstrate that AI systems are secure, safe and trustworthy.

Explore IAIST → Join the Alliance