MA S37: Promoting economic development with emerging artificial intelligence models and safety

In one paragraph

Massachusetts' S 37 would create an Artificial Intelligence Innovation Trust Fund for grants to AI developers and research, and a new Chapter 93M on frontier AI models. Developers of covered models (trained with more than 10^26 operations costing over $100 million) would need safety protocols, annual third-party reviews and 72-hour incident reports to the attorney general. The committee reported a new draft, S 2630, instead of S 37 on October 16, 2025; S 37 itself did not advance.

Plain-English summary, written with AI assistance. The official text controls.

What it means for business

None yet: the bill did not advance in this form. Developers of very large AI models and operators of large computing clusters would have faced safety protocols, audits, incident reporting and penalties.

Key provisions

1. 1Would create the Massachusetts Artificial Intelligence Innovation Trust Fund for grants to companies developing or deploying AI models, entrepreneurship programs and research
2. 2Covered model: trained with more than 10^26 operations at a cost over $100,000,000, or fine-tuned from one with at least 3x10^25 operations at a cost over $10,000,000 (adjusted for inflation); from January 1, 2027 the compute thresholds would be set by the secretary of technology services and security
3. 3Before training, developers would have to implement cybersecurity protections, a full-shutdown capability and a written safety and security protocol, publish a redacted copy and send it to the attorney general
4. 4Developers would have to assess critical-harm risk before use or release, retain an annual independent third-party investigator, and send the attorney general an annual compliance statement
5. 5Developers would have to report each AI safety incident to the attorney general within 72 hours
6. 6Computing-cluster operators would have to collect customer identity and payment information and be able to shut down resources used to train a covered model
7. 7Employees of developers would be protected when disclosing risks or violations; developers would need an anonymous internal reporting process
8. 8The attorney general would enforce, with civil penalties up to 5% (first violation) or 15% (later) of training compute cost for violations causing harm or an imminent public-safety risk, and up to $25,000/$50,000 for investigators and cluster operators

Latest legislative action

Oct 16, 2025. Accompanied a new draft, see S2630

Sponsors (showing 5 of 48)