Skip to main content

Standards & Control Mapping

Mapped to Tests and Evidence — Never Certified.

How Mythos AI Security findings relate to security and AI frameworks, and exactly where mapping ends and audit responsibility begins.

Three different things

Mapping, reference, certification.

Only the first is something Mythos AI Security does.
  1. Mapped

    Findings are tagged to the controls they touch, in four frameworks plus vertical packs. Evidence of gaps — self-reported, not audited.

  2. Referenced

    Frameworks scoping draws on so evidence can support your own reviews. No mapping is claimed.

  3. Certified

    Not offered. Mythos AI Security does not certify, attest or declare compliance with any framework.

Mapping built

Four frameworks, touched not satisfied.

A mapped control is one a finding touches — never one Mythos AI Security marks as satisfied. The map shows evidence of gaps. Separately, Minotaur maps its own benchmarks to 15 standards and 86 clauses, with no badges.

  • NIST SP 800-53Mapping builtSecurity and privacy controls.
  • OWASP Top 10 (2021)Mapping builtWeb application risks.
  • OWASP Top 10 for LLM Applications (2025)Mapping builtRisks specific to applications built on language models.
  • DoD Zero TrustMapping builtZero-trust reference architecture and strategy capabilities.
  • Vertical packsMapping builtHealthcare, financial services, federal and general AI, layered on the four above.

Referenced in scoping

Frameworks we draw on.

Named so evidence can be organized for your reviews. Not mapped, aligned or certified.

AI risk and management

  • NIST AI Risk Management Framework and its Generative AI Profile
  • ISO/IEC 42001

Adversarial and agent threats

  • NIST AI 100-2 adversarial ML taxonomy
  • OWASP guidance for agentic applications
  • MITRE ATLAS

Secure development and response

  • NIST SP 800-218 / 218A (SSDF)
  • NIST SP 800-61r3 (incident response)

Sector and public sector

  • US Treasury financial-services AI risk framework
  • NYDFS Insurance Circular Letter No. 7
  • OMB M-25-21 and M-25-22
  • NIST critical-infrastructure AI profile
  • EU Cyber Resilience Act

Formats in use

Open formats, in use.

Receipts, signatures and traces use these published formats today. A policy adapter for OPA or Cedar is planned, on customer demand.
  • DSSE envelopes
  • Ed25519 signatures
  • Merkle inclusion proofs
  • OpenTelemetry GenAI conventions
  • ISO 4217 currencies
  • OPA or Cedar policy adapter(planned)

Not listed, on purpose

SOC 2, ISO/IEC 27001, PCI DSS, HIPAA, HITRUST, GDPR, CMMC and FedRAMP are not mapped. A supplier’s SOC 2 report can be recorded as a vendor-asserted claim, which caps at supported — never as a Mythos AI Security result.

Mythos AI Security

Evidence your reviewers can map.

Tell us which frameworks your reviewers use, and we will show how findings would map to them.