AI Frontier Institute

AI Innovation Trust Fund

View primary source →

North Carolina Senate Bill 735 (2025) has not been enacted. Its last recorded action was re-referral to the Appropriations/Base Budget committee on March 26, 2025. If enacted, it would add a new Part 18A to Article 10 of Chapter 143B of the General Statutes. The Part would create the North Carolina Artificial Intelligence Innovation Trust Fund (§ 143B-472.83A). The Secretary of Commerce would be trustee. The fund would pay for grants or financial assistance to companies developing or deploying AI models in key sectors, and for AI entrepreneurship programs. Fund money could not support AI for unconstitutional mass surveillance, unlawful social scoring, discriminatory profiling, or deceptive content meant for fraud or electoral interference. The Secretary could convene an AI Innovation and Safety Advisory Panel and could consider NIST and EU AI Act standards. The bill would also set a safety regime for developers of "covered models," meaning models the Secretary identifies as warranting oversight (§ 143B-472.83B). Before training, developers would need to: implement cybersecurity protections; be able to promptly shut down the model; and keep a written safety and security protocol, publish a redacted version, and give a copy to the attorney general. Before use or release, developers would need to assess whether the model could cause or enable "critical harm," meaning mass-casualty weapons, attacks on critical infrastructure causing at least $500,000,000 in damage, or comparable harms. Developers would also need to retain testing records. They would have to hire an independent third-party investigator each year, submit an annual compliance statement signed by the chief technology officer or a more senior officer, and report AI safety incidents to the attorney general within 72 hours. Several provisions on tiered compliance, misuse responsibility, fairness assessments, and AI-interaction disclosure are phrased as optional ("may"). Operators of large computing clusters would have to collect customer identity and payment information, assess whether the customer intends to train a covered model, keep logs and records for seven years, and be able to shut down resources (§ 143B-472.83C). The attorney general would enforce the Part (§ 143B-472.83D). Civil penalties for developers could reach 5% of training compute cost for a first violation and 15% for later ones. Penalties for investigators and cluster operators would be up to $25,000 for a first offense, $50,000 for later violations, and $5,000,000 in aggregate for related violations. Contract clauses that waive or shift liability would be void, and affiliated entities could face joint and several liability if they structured themselves to avoid liability. Whistleblower protections would bar developers from preventing or retaliating against employee disclosures. They would also require employee notice and an anonymous internal reporting process (§ 143B-472.83E). The Secretary would file an annual report on the AI workforce, issue regulations, update compute thresholds and investigation requirements each year, and issue guidance (§ 143B-472.83). Section 2 would appropriate $750,000 nonrecurring from the General Fund to the Department of Commerce for 2025-2026. Section 3 would make the act effective July 1, 2025.

← Back to the tracker