Three of the biggest names in AI are squaring off over a Massachusetts bill that could become the toughest state-level AI safety law in the country, and they’re not on the same side.

Anthropic is actively championing Massachusetts’ proposed AI regulations, calling the bill “the strongest state AI safety proposal in the country.” OpenAI and Google are pushing back hard, arguing the requirements would be too burdensome. It is a notable crack in what many assumed was a united tech industry front on AI governance.

The legislation, currently in bicameral negotiations between the Massachusetts House and Senate, targets the largest AI labs, specifically those generating more than $500 million in AI-derived revenue or spending more than $1 billion on AI research and development. If passed, it would require those companies to submit to independent risk evaluations at least every six months.

What the Bill Would Actually Require

Both chambers agree on the high-level goal: preventing catastrophic AI risks by putting guardrails around frontier model development. But they diverge on enforcement, scope, and how much companies would have to open their systems to external scrutiny.

The stricter elements Anthropic is supporting would require AI labs to demonstrate that their models cannot produce outputs that could lead to mass harm events, and to have those claims independently verified. Critics from OpenAI and Google argue these requirements are poorly defined, difficult to operationalize, and would slow innovation without meaningfully improving safety outcomes.

OpenAI has said publicly it wants to align the Massachusetts proposal with laws already on the books in states like Illinois, which impose lighter transparency requirements. OpenAI is not opposed to AI safety legislation in principle. It is opposed to this specific bill as written.

The Deeper Split

What makes this story more than a standard lobbying battle is what it reveals about how these companies actually think about AI risk.

Anthropic’s public position has consistently been that frontier AI models carry serious risks that market incentives alone will not address. Supporting state-level regulation, even if it creates more friction for competitors operating under looser frameworks, fits with that worldview.

OpenAI and Google have tended to favor federal preemption: one consistent national standard rather than a patchwork of state laws each with their own requirements and enforcement agencies. The Massachusetts fight is partly about who gets to set that standard.

Anthropic made campaign contributions exceeding $120,000 to Massachusetts Democrats in August 2026, suggesting the company views this as a meaningful fight worth investing in politically.

What This Means for Businesses Using AI

If the stricter version passes, the immediate impact falls on the AI labs themselves, not the businesses buying their APIs and models. But there are ripple effects worth tracking.

Compliance overhead moves upstream. Laws that require AI developers to independently verify model safety claims will likely result in more conservative model releases, longer evaluation periods, and potentially higher costs passed downstream. Businesses that rely on cutting-edge frontier models for competitive advantage should expect slower upgrade cycles if the Massachusetts framework spreads.

Documentation requirements could follow. Several versions of the bill include provisions that would require developers to maintain detailed records about training data, safety testing, and model behavior. This kind of provenance tracking often becomes a requirement for enterprise procurement, even when not mandated by law, once regulators start asking for it.

State-by-state variation creates compliance complexity. If California, Massachusetts, and other states each pass different AI safety frameworks, businesses operating nationally will need to track which AI vendors are compliant in which jurisdictions. That is a real operational headache, especially for mid-market companies without dedicated legal and compliance teams.

The Anthropic stance may reshape enterprise procurement. Anthropic championing stricter safety standards, and the willingness to break publicly from peers to do it, will resonate with buyers in regulated industries like healthcare, finance, and legal. If the narrative becomes “Anthropic chooses safety over growth while competitors push back,” that will influence enterprise vendor decisions.

Where This Goes Next

The bill is in bicameral conference negotiations as of late August 2026. The final text could look substantially different from either chamber’s version. Governor Maura Healey has not publicly committed to signing it in any form.

Regardless of what passes in Massachusetts, this fight signals a broader shift underway: AI regulation is no longer a theoretical future concern. State legislators are actively writing enforceable law, AI companies are lobbying aggressively, and the dividing lines are getting clearer.

The companies that will be best positioned, as vendors and as buyers, are those that are building compliance readiness now, not scrambling to interpret new requirements after they pass.

---

For businesses evaluating AI adoption strategies, Enterprise DNA’s advisory services help you build AI-powered workflows that account for regulatory requirements as well as operational outcomes. Book a session with our team to talk through what responsible AI adoption looks like for your industry.