The Standard
The 841 Dimensions Regulatory Mapping Version History Assurance Framework Scoring Methodology
Compliance
Compliance Leaderboard Compare Platforms
Verification
Verification & Pricing Request Assessment
About
About AgentGoverning Press & Media
Contact
Version History

Standard Version History

AGS is a living standard; v3.1 expanded it to 841 dimensions. Amendments are published following a public comment period.

Changelog

Release Timeline

AGS v3.1 — June 2026 Current
  • Expanded to The 841 Dimensions (+3 cross-border dimensions)
  • Cross-border completion: law-following AI (compliance-by-design), jurisdictional compliance kill-switch, agent contract-formation authority
  • Provable coverage maintained across 12 framework taxonomies (113 threats), 0 gaps
  • v3.1.1 patch (independent review): citation corrections to NIST AI RMF GOVERN subcategory and EU AI Act article references across the new dimensions; no dimension changes (841 / 101 unchanged)
AGS v3.0 — June 2026
  • Expanded to The 838 Dimensions across 101 landscapes (+21 dimensions, +1 landscape)
  • Group K — Frontier Autonomy: scalable oversight of superhuman systems, recursive-self-improvement governance, untrusted-model control & compute governance, long-horizon autonomy limits, agent-economy systemic risk, and model welfare under moral-status uncertainty
  • Group L — Embodied AI: safety-class & force/speed limiting, physical-action reversibility & fail-safe-to-stop, sim-to-real/SOTIF validation, robot-fleet & OTA-update governance
  • Aligned to frontier-lab safety frameworks, UK AISI research, and ISO robot-safety standards; provable coverage maintained across 10 framework taxonomies
AGS v2.2 — June 2026
  • Expanded to The 817 Dimensions across 100 landscapes (+21 dimensions, +6 landscapes)
  • Frontier-safety evaluation: deceptive-alignment/scheming, sabotage, corrigibility, sandbagging, capability-threshold gating, dangerous-capability elicitation
  • Non-human identity lifecycle, agent compute & cost budgeting, code-execution sandboxing, autonomous payment mandates
  • New safety-critical sector landscapes: aviation, maritime, space, nuclear, gambling, sports integrity
  • Provable coverage: 96 threats across 10 framework taxonomies (OWASP, NIST AI 600-1, MITRE ATLAS, CSA MAESTRO, NHI), 0 gaps
AGS v2.1 — April 2026
  • Initial publication of The 796 Dimensions
  • 10 dimension groups (A through J)
  • First public compliance leaderboard
  • 1 platform verified · 5 platforms on leaderboard
  • 180,767 words of specification
Roadmap — Beyond v3.0 Planned
  • Continued frontier tracking: scalable-oversight, AI-control, and compute-governance controls as the research matures
  • Deeper cross-border & sovereignty controls (expanding AG-47) and further sector coverage
  • Living standard: amendments published following a public comment period
Process

How to Contribute

Proposed amendments are submitted to the standards committee for review. Submissions are published for a 90-day public comment period before adoption. No amendment is adopted without independent review.

framework@agentgoverning.com
Independence

Conflict of Interest Policy

No commercial product or vendor may serve on the committee that governs this standard. AgentGoverning accepts no commercial relationships that could influence standards or scores.