LETTER: Massachusetts should follow Illinois’s lead without going further

Dear Members of the [Mass. Senate] Conference Committee:

On behalf of Chamber of Progress, a tech industry association supporting public policies to build a society in which all people benefit from technological advances, I respectfully urge you to oppose S. 3178, unless amended. While the current Senate bill represents an improvement over the House version, its frontier AI provisions should be further amended to mirror Illinois SB 315 and remove the requirement that developers undergo independent model evaluations. These changes would provide meaningful safety and transparency requirements while ensuring that Massachusetts adopts a framework that regulators and covered developers can realistically implement.

Massachusetts should follow Illinois’s lead without going further

The current Senate bill represents an improvement over the House version, but Massachusetts should further amend it to mirror Illinois SB 315. California’s SB 53 established the initial state framework for frontier AI regulation, and New York subsequently amended its RAISE Act to more closely align with that approach. Illinois then built on those laws through SB 315, creating a detailed framework for safety and security protocols, transparency reporting, critical safety incidents, whistleblower protections, and independent compliance audits. Because all three laws are still in the early stages of implementation, lawmakers do not yet have evidence showing how regulators will interpret key definitions, whether the reporting requirements will produce useful safety information, or what practical challenges may arise during compliance.

Following Illinois would allow Massachusetts to rely on a comprehensive framework that has already been debated and enacted, while giving regulators clearer definitions, established obligations, and a more workable structure to administer. Massachusetts does not need to go beyond Illinois to establish meaningful oversight, particularly before any of these new state frameworks have completed their first implementation cycle.

Massachusetts should not mandate independent model evaluations before reliable standards and capacity exist S. 3178 would require covered developers to submit their frontier models to independentevaluations for catastrophic capabilities and risks at least once every 120 days. Yet the bill imposes this mandate without an established accreditation system for determining who is qualified to perform these evaluations, a broadly accepted testing methodology, or objective benchmarks for determining whether a model has passed. Catastrophic-risk evaluations can require specialized technical expertise, secure access to sensitive model information, and extensive testing across multiple capabilities and potential threat scenarios. Without clear standards, different evaluators could reach inconsistent conclusions about the same model, leaving developers uncertain about what the law requires and regulators without a reliable basis for comparing results.

The 120-day schedule creates an additional problem because frontier models and their safeguards are continuously updated. An evaluation could be outdated by the time it is completed, forcing developers into overlapping review cycles that consume significant technical resources without providing an accurate assessment of the model currently in use. Massachusetts should remove this requirement rather than impose an enforceable mandate that cannot yet be administered consistently or reliably.

For these reasons, I respectfully urge you to oppose S.3178, unless amended. Massachusetts can advance meaningful frontier AI oversight without imposing requirements that lack the standards, capacity, and infrastructure necessary for reliable implementation.

Sincerely,

Jabari Cooper

Director of State and Local Government Relations, Northeast

Chamber of Progress