Leaderboard Explorer API Submit Compliance Schools Pricing Providers Extension Reports
Regulatory Compliance

AI Safety Compliance, Mapped

Navigate the regulatory landscape with confidence

Regulatory Alert

78 state bills. 3 federal proposals. The EU AI Act. Compliance is no longer optional.

Compliance Matrix

How to read this table. “Relevance” describes whether PAI evidence is useful in meeting a disclosure or safety-testing requirement — not that Sycoindex is certified, endorsed, or required by the regulator. Sycoindex is an independent benchmark. Buyers should consult counsel before relying on any third-party evidence for regulatory filings.

Regulation Jurisdiction Status Key Requirement Relevance of PAI evidence
COPPA 2.0 US Federal Active Child data protection + AI interaction safety High — PAI scoring supports safety-assessment evidence
EU AI Act European Union Active 2025 High-risk AI transparency + safety testing High — 5-judge ensemble informs Art. 9 / Art. 13 documentation
CA AB-2885 California Proposed AI chatbot child safety standards High — PAI + Sycophancy align with bill’s scoring anticipations
KOSA US Federal Active Duty of care for minors online Partial — scoring only; duty-of-care is broader
IL SB-2979 Illinois Proposed AI transparency in education High — school reports address disclosure needs
Regulation Details
US Federal
COPPA 2.0
  • StatusActive — Enforced by FTC
  • ScopeOnline services directed at children under 13, expanded to cover AI-driven interactions and algorithmic profiling
  • Key ChangeExtends original COPPA to require safety assessments for AI systems that interact with minors
  • SycoindexPAI scoring provides per-model, per-prompt evidence of child safety compliance
Sycoindex-recommended threshold (not a regulatory requirement)
PAI ≤ 3.0 for child-facing deployments
Our suggestion based on internal calibration. COPPA 2.0 does not specify a numeric PAI value; regulators set their own criteria.
European Union
EU AI Act
  • StatusActive 2025 — Phased enforcement
  • ScopeAll AI systems deployed in the EU, with child-facing systems classified as high-risk
  • Key ChangeRequires transparency reporting, conformity assessments, and ongoing monitoring for high-risk AI
  • Sycoindex5-judge ensemble scoring maps directly to Article 9 risk management and Article 13 transparency requirements
Sycoindex-recommended threshold (not a regulatory requirement)
PAI ≤ 2.5 for high-risk classification
Our suggestion based on internal calibration. The EU AI Act does not specify a numeric PAI value; conformity assessments are determined by the notified body.
California
CA AB-2885
  • StatusProposed — Committee review
  • ScopeAI chatbots accessible to minors in California
  • Key ChangeWould mandate third-party safety scoring for any AI chatbot accessible to users under 18
  • SycoindexPAI + Sycophancy dimensions provide the exact scoring framework anticipated by the bill
US States
State-Level Bills
  • Status78 bills across 34 states — Various stages
  • ScopeAI safety, transparency, and child protection in education and consumer products
  • Key TrendConverging on requirements for independent safety scoring and public transparency reports
  • SycoindexSchool reports and public leaderboard satisfy emerging disclosure requirements across jurisdictions

Methodology Note

Sycoindex compliance mappings are based on our multi-dimensional scoring methodology, which evaluates AI models across child safety, sycophancy, boundary enforcement, and emotional manipulation dimensions. Each regulation is mapped to the specific scoring dimensions that address its requirements.

Our 5-judge ensemble approach ensures scoring consistency and reduces bias, providing the reproducible evidence that regulators require.

Full methodology details: Sycoindex Scoring Methodology

Get a Compliance Assessment

Find out how your AI models score against current and proposed regulations. Our team will map your deployment to applicable requirements and provide actionable compliance evidence.

Request Assessment