Skip to main contentSkip to navigation
CSOAI
Home
Verify a card
Reference / archiveA dated reference page, kept for the record. For the current experience, start at the measurement board.Academy & Training →
Dashboard
AI Systems
Compliance
Measurement Council
The Watchdog
Reports
Training
Governance Workbench
API Docs
PDCA Cycles
Knowledge Base
Regulator View
Leaderboard
Recommendations
Standards
Resources
About
CSOAI

Unable to load review

The exam attempt could not be found.

Models refuse generic harm. They do not refuse the regulated practices.

GSPC board — 13 measured of 14, plus jail/containment (18 Aug stamp, n=71, separation untested). Signed stamp; live axis and model counts at /api/gspc. Harness published — recompute it.

Stay Updated on AI Safety

Get the latest insights on AI regulations, safety frameworks, and industry updates.

CSOAI

Building the future of AI safety through independent training, signed attestation, and transparent measurement.

Measurement, not certification — every published finding links to a signed, recomputable record.

Product

  • Training Courses
  • Training records (we certify nothing)
  • Watchdog Reports
  • Analyst Workbench
  • Global AI Regulation Tracker
  • Framework Crosswalk (13×8)
  • Library — full archive
  • AI Glossary
  • Blog
  • FAQ

Evidence & APIs

  • GSPC board — live JSON
  • Regulation feed — live JSON
  • Corrections feed — live JSON
  • Verify a card
  • llms.txt — for answer engines
  • did:web trust root (did.json)
  • API Documentation
  • Methodology
  • The honesty gate — our own losses
  • Meta-benchmark index — other benchmarks, cited
  • Firewall Charter — measure, never fix

Company

  • About Us
  • Technology
  • Integrations
  • Partners & Advisory
  • Trust Center
  • Case Studies
  • Status
  • Careers
  • What we do NOT accredit
  • Contact

Legal

  • Important Disclaimers
  • Privacy Policy
  • Terms of Service
  • Cookie Policy
  • Data Processing (GDPR)
  • Service Level Agreement
  • Insurance Certificate (on request)
Library — full archiveEU AI Act & RegulationRegions & JurisdictionsAcademy & TrainingLayer-0, MCP & VerificationGSPC Axes & BenchmarksGovernance & FrameworksProduct, OS & DemoCompany, About & Legal
FrameworksEU AI ActArticle 50 — transparencyNIST AI RMFISO 42001DORAAgent Governance

© 2026 CSOAI. All rights reserved.

Privacy PolicyTerms of ServiceMembershipLicensing

Verify us everywhere

  • GitHub · CSOAI-ORG
  • Hugging Face · csoai
  • PyPI · inspect-signed-receipt
  • PyPI · csoai
  • PyPI · proofof-ai-mcp
  • Kaggle
  • Zenodo DOI 10.5281/zenodo.21991104
  • Wikidata · Q141128616
  • Companies House · 16939677
  • did:web trust root

Every listing above is a live, independently hosted record — follow any of them to check us.

Standards participation & memberships

Decentralized Identity Foundation (DIF) — participant·Content Authenticity / C2PA — Contributor (conformance in progress)·Zenodo DOI 10.5281/zenodo.21991104

This site uses AI systems, including the Council assistant. Every AI surface is disclosed at first interaction under EU AI Act Article 50 and classified publicly on /ai-transparency.

Human oversight applies to every governed action on this platform (Article 14): measurements are machine-run, judgements are human-owned. Our public artefacts carry signed provenance (Ed25519-signed records; C2PA conformance in progress — see /provenance-finding).

CSOAI is an independent organization with no financial ties to OpenAI, Anthropic, Google, Microsoft, Meta, or any AI vendor. Our only incentive is public safety and workforce development.

Council of AI — CSOAI Ltd, UK Companies House 16939677, London. Professional Indemnity Insurance up to £5,000,000 (policy number on request). Contact: [email protected].

Essential cookies only by default; anonymous analytics need consent. Details