Independent authority. No vendor affiliation. Certification cannot be purchased.
AgentGoverning is an independent, non-commercial authority publishing governance standards for autonomous AI agents.
Agent Aegis comprises two independent audit types: the LLM Audit, which evaluates foundation model governance across 792 dimensions, and the Agent Audit, which assesses autonomous agent deployments against the same standard. Both audits employ independent LLMs as adversarial attack generators — GPT-4o, Gemini 2.5 Flash, and Grok-3 for the verified benchmark run — to ensure adversarial rigour. Agent Shield™ is the first platform to complete both audits.
We publish standards. We verify claims. We publish results. We do not sell products. We do not endorse vendors. We accept no commercial relationships that could influence our standards or our scores.
AgentGoverning receives no vendor funding. The AGS standard is published under Creative Commons BY 4.0. Any platform may submit for assessment. The standard is the standard — not a product.
The 792 Dimensions of AI Agent Governance were developed through analysis of emerging AI agent deployments, regulatory frameworks across multiple jurisdictions, and adversarial testing methodology.
This standard is versioned using semantic versioning. Breaking changes increment the major version number.
Proposed amendments are published for a 90-day public comment period before adoption.
No commercial product or vendor may serve on the committee that governs this standard.
Certification against this standard is issued by AgentGoverning following independent verification. Certification is not self-assessable. Certification cannot be purchased. Results are published publicly regardless of outcome.
AI agents are executing financial transactions autonomously. Without structural governance, your exposure is unlimited. AGS v2.1 defines the financial containment controls that every agent deployment requires — from mandate enforcement to transaction structuring detection.
See value governance dimensions →Traditional security tools were built for predictable software. AI agents are non-deterministic. AGS v2.1 defines the security and identity dimensions specifically designed for autonomous agents — from cryptographic attribution to adversarial coordination detection.
See security dimensions →SOX, GDPR, EU AI Act, and FCA all have implications for AI agent deployments. AGS v2.1 maps every dimension to its regulatory obligation. No other framework does this for autonomous agents specifically.
See regulatory mapping →The question is not whether to deploy AI agents. It is whether you can demonstrate they were governed. AGS v2.1 provides the independent Agent Aegis benchmark for that demonstration — the standard your auditors, regulators, and insurers will reference.
See certification tiers →