Policy Briefing — NTIA Releases AI Accountability Policy Report

Executive briefing: On 27 March 2024 the National Telecommunications and Information Administration (NTIA) released the AI Accountability Policy Report, responding to Executive Order 14110. The report outlines policy recommendations to ensure independent evaluations of high-risk AI systems, clearer disclosures to users, and improved incident reporting to regulators.

Key recommendations

Independent evaluations. Developers and deployers of high-risk AI should undergo third-party audits and certifications, with evaluators adhering to professional standards.
Transparency disclosures. NTIA urges mandatory system cards, provenance documentation, and disclosure of limitations to downstream users.
Incident reporting. Organisations should establish channels to report AI failures, bias, and security events to regulators and affected users.
Regulatory coordination. Federal agencies should harmonise sectoral rules, referencing NIST AI Risk Management Framework and sector-specific standards.
Liability alignment. NTIA supports clarifying liability for AI harms and ensuring that contractual terms do not hinder accountability.

Program actions

Audit readiness. Prepare evidence packs, testing documentation, and governance artefacts aligned with anticipated audit criteria.
Disclosure tooling. Build system cards and provenance reports that explain training data, evaluation metrics, and limitations for customers and regulators.
Incident escalation. Align AI incident response with cybersecurity and safety processes to capture bias, misuse, and failure events.
Contract review. Update vendor and customer contracts to reflect accountability expectations, including audit rights and remediation commitments.

Sources

Zeph Tech is translating NTIA’s accountability recommendations into audit, transparency, and incident response programmes for U.S. enterprises.