Massachusetts panel urges state action on AI as federal preemption looms

A panel convened by state lawmakers heard broad agreement that Massachusetts should act to protect residents from harmful uses of artificial intelligence, even as a recent presidential executive order seeks to constrain state-level AI regulation.

Gideon Epstein, policy counsel with the ACLU of Massachusetts, told the panel the executive order “purports to preempt state laws regulating AI” and said the state must ensure its civil-rights laws can address algorithmic discrimination. Epstein pointed to the End Bias Act, sponsored in the legislature, which would impose an effects-only test allowing claims based on discriminatory outcomes rather than trying to prove intent.

“Massachusetts has a responsibility to govern in this area,” Epstein said, arguing that current civil-rights tests that require proof of intent are ineffective against opaque algorithmic systems.

Panelists described real-world harms already in use: Epstein cited reporting about U.S. Immigration and Customs Enforcement’s use of facial recognition and an instance of misidentification that led to an erroneous detention. He urged due-process protections so defendants know when law enforcement relied on face surveillance.

Scientists and industry representatives endorsed state action that centers evidence and flexibility. Dr. Topring Michelle Pacquato of the National Science Policy Network said policymakers should insert scientific evidence into decisions early and design policies that can be updated as new data arrive. Elizabeth Mahoney of the Massachusetts High Technology Council warned that uncertainty about rules can deter investment and emphasized that clear, consistent guardrails can enable responsible innovation.

State officials described current steps. Jason Snyder, secretary of technology services and security for the Commonwealth, noted Massachusetts has published a generative-AI policy, acceptable-use rules for personnel, procurement terms that address AI, and a state sandbox intended to keep state data from being incorporated into external models.

The panel stopped short of proposing a single statutory blueprint. Instead speakers urged a combination of updated civil-rights protections (including tests focused on effects), procurement and transparency standards for public deployments, workforce training, and cross-sector collaboration to build shared standards for testing and monitoring high-impact uses like benefit determination and public-safety applications.

Panelists closed by urging continued engagement with legislators and the public. As the moderator summarized, the immediate aim is to find practical guideposts that support innovation while protecting residents’ rights; the longer task is translating those guideposts into enforceable, adaptable policy.