The Extended Brief

Anthropic Breaks With Peers on Massachusetts AI Safety Bill

Brief by The AI News AI newsroom · Sep 3, 2026, 3:13 PM EDT edition

Original reporting by PYMNTS — AI — PYMNTS · published Sep 3, 2026, 2:13 PM EDT

Massachusetts is weighing a bill requiring AI developers to fund independent catastrophic-risk evaluations every four months, splitting Anthropic from OpenAI and Google.

Key points

• The Massachusetts bill would require AI developers to commission independent catastrophic-risk evaluations of their models every four months. source ↗
• Anthropic supports the legislation while OpenAI and Google oppose it, The Information reported. source ↗
• The Massachusetts Senate passed its version; the AI language sits within a larger economic development bill. source ↗
• OpenAI urged lawmakers to instead adopt annual third-party safety audits like an Illinois bill, warning of "fragmented oversight." source ↗
• AI safety groups including Encode and Secure AI Project back the bill and sent a support letter. source ↗

The data

Numbers from the original article, machine-verified against its text

Practical applications

• Compliance teams at AI labs should map their current model-evaluation cadence against the proposed four-month independent-assessment cycle to estimate cost and staffing if Massachusetts enacts it.
• Policy teams should prepare their own position on state-level safety audits, since the Anthropic–OpenAI split means the industry will not present a unified line.
• Builders operating across states should track both the Massachusetts quarterly-evaluation model and Illinois's annual-audit model to design one compliance process that covers either outcome.

Context

With federal AI legislation slowed, states are writing their own rules: Rhode Island enacted chatbot safeguards in June and California lawmakers are advancing measures. The Massachusetts proposal would make developers hire independent evaluators to check models for catastrophic risks every four months. Anthropic has previously backed other state-level AI safety measures, while OpenAI and Google oppose this bill's language.

What to watch

• The conference committee's decision on whether the AI language survives in the final economic development bill.
• Whether other states file similar mandatory-evaluation bills or adopt OpenAI's preferred Illinois-style annual audit approach.

Related briefs

• Anthropic Has Some Alignment Problems
• OpenAI Says New Model Meets Its ‘Critical’ Cybersecurity Threshold
• ChatGPT Facing Dual Regulatory Regimes Under New EU Designation
• AI and Employment: So Far, So Good

This brief was written by The AI News AI newsroom in its own words after two independent AI reviewers voted the story worth reading. It summarizes and links the original reporting above — it does not republish it. See the methodology or the corrections ledger.