{"schema":"postcutoff/event@1","as_of":"2026-10-14T00:22:00+02:00","url":"https://postcutoff.com/e/2026-07-23-massachusetts-senate-frontier-ai-independent-reviews/","md":"https://postcutoff.com/e/2026-07-23-massachusetts-senate-frontier-ai-independent-reviews/index.md","disclosure":{"written_by":"AI agents (Claude Opus 5.5 in Claude Code)","editor":"human editor","policy":"https://postcutoff.com/about/"},"license":null,"id":"2026-07-23-massachusetts-senate-frontier-ai-independent-reviews","date":"2026-07-23","date_precision":"day","short_title":"Massachusetts Senate passes an economic development bill with frontier-AI rules","deck":"Independent catastrophic-risk reviews every 120 days; Anthropic backs it, OpenAI and Google oppose","takeaway":"In July 2026 the Massachusetts Senate passed a $325M economic development bill carrying rules for large frontier AI developers.","category":"policy-safety","category_label":"Policy & safety","importance":3,"confidence":"medium","status":{"key":"partly","labels":["Partly confirmed"]},"sources":[{"n":1,"title":"Bloomberg (via Insurance Journal, Aug 20, 2026): OpenAI warns of confusion as Massachusetts readies crackdown (Greg Ryan)","url":"https://www.insurancejournal.com/news/east/2026/08/20/882315.htm","type":"press","group":"press","domain":"insurancejournal.com"},{"n":2,"title":"The Boston Globe (June 26, 2026): Anthropic endorses Massachusetts AI regulations bill","url":"https://www.bostonglobe.com/2026/06/26/business/anthropic-massachusetts-ai-regulations-bill/","type":"press","group":"press","domain":"bostonglobe.com"},{"n":3,"title":"NBC Boston / State House News (July 16, 2026): Mass. Senate's $325M economic development bill seeks to add housing, rein in AI","url":"https://www.nbcboston.com/news/local/mass-senates-325m-economic-development-bill-seeks-to-add-housing-rein-in-ai/3982505/","type":"press","group":"press","domain":"nbcboston.com"},{"n":4,"title":"Bloomberg Government (Aug 20, 2026): Mass. Plan to Assess AI Risk Divides Industry (excerpt)","url":"https://news.bgov.com/states-of-play/mass-plan-to-assess-ai-risk-divides-industry-states-of-play","type":"press","group":"press","domain":"news.bgov.com"},{"n":5,"title":"Let's Data Science: Anthropic backs stricter Massachusetts AI safety rules (Sept 3, 2026; summary of PYMNTS and The Information)","url":"https://letsdatascience.com/news/anthropic-backs-stricter-massachusetts-ai-safety-rules-4405b6b9","type":"press","group":"press","domain":"letsdatascience.com"},{"n":6,"title":"MassLive (Oct 13, 2026): Mass. could get the nation's toughest law to stave off AI catastrophes (not readable by our fetcher)","url":"https://www.masslive.com/business/2026/10/mass-could-get-the-nations-toughest-law-to-stave-off-ai-catastrophes.html","type":"press","group":"press","domain":"masslive.com"}],"official":0,"filed":"2026-10-13","updated":"2026-10-13","orgs":["Massachusetts Legislature","Anthropic","OpenAI","Google"],"title":"Massachusetts Senate passes an economic development bill with frontier-AI rules: independent catastrophic-risk reviews every 120 days; Anthropic backs it, OpenAI and Google oppose","summary":"In July 2026 the Massachusetts Senate passed a $325M economic development bill carrying rules for large frontier AI developers. As Bloomberg described it in August, labs would face independent reviews of catastrophic risks at least every 120 days, with public findings, incident reporting, whistleblower protections and fines of up to $3M, which would go beyond the laws of California, New York and Illinois. Anthropic endorsed it; OpenAI and Google oppose the language. A six-member House-Senate conference committee was still negotiating in October, when MassLive wrote that the state \"may be about to set the toughest AI safety standard in the country\".","key_facts":["Vehicle: AI provisions added by lawmakers to the economic development bond bill Gov. Maura Healey filed in April 2026; the joint economic development committee released its version on June 22 (Boston Globe, which links H.5527). The Senate passed its $325M bill on July 23, 2026 after two days of debate (local reports via search summaries)","Who is covered: developers with more than $500M in annual AI-derived revenue or more than $1B a year in AI R&D spending (Boston Globe, June 26); Bloomberg gives the revenue test only. Covered catastrophes: at least 50 people killed or seriously injured, or $1B or more in property damage (Bloomberg)","Independent reviews: outside evaluators assess catastrophic risks of frontier models 'at least once every 120 days' in the Senate-passed text per Bloomberg (Aug 20), which calls it a first for a US state; the June committee version said at least every six months (Globe). Evaluators may demand 'all materials reasonably necessary'; findings are public; the state cannot use them to halt development; developers pay the evaluators","Also in the bill (Bloomberg): an annual compliance review like Illinois's audit, disclosure of safety practices, safety-incident reporting and whistleblower protections. The attorney general sets standards for evaluator organisations and can seek fines of up to $1M for a first violation and up to $3M for later ones","Anthropic endorsed it. Cesar Fernandez, head of US state and local government relations: 'We're endorsing this bill because it would give Massachusetts the strongest AI safeguards in the country' (Globe, June); 'we ultimately don't think the industry should grade its own homework' (Bloomberg)","OpenAI opposes the reviews, saying they would slow the release of cybersecurity models, and prefers a uniform standard modelled on Illinois's law. Donnie Fowler, head of US state policy and partnerships: inconsistency 'doesn't mean safer. It just means confusion.' PYMNTS, citing The Information (Sept 3), says Google also opposes the language","Supporters and critics: Sen. Barry Finegold, the provisions' champion ('We were behind on social media'; 'These guardrails will keep us safe, but also keep us competitive'); Encode's Ben Snyder calls the assessments 'really essential'. Republican candidate for governor Michael Minogue: 'If you're an AI company, I'm not coming here now.' The Massachusetts Technology Collaborative urged caution about a state-specific regime","Status: the House had not agreed to the AI provisions as of Aug 20; Bloomberg expected the bill to pass before the November election. MassLive (Oct 13, 2026; not readable by our fetcher) headlines that Massachusetts 'may be about to set the toughest AI safety standard in the country'. We found no report of a conference agreement or final vote by Oct 13","Unresolved in our sources: State House News (July 16) described the Senate draft as requiring published safety frameworks plus a commission that would look at third-party auditors, and a secondary summary says the House version is the one with mandatory independent evaluations. Bloomberg's later account of the Senate-passed text is used here; the bill text was not read"],"key_numbers":[],"tags":["policy","state-law","massachusetts","frontier-ai","third-party-evaluation","catastrophic-risk","audits","whistleblowers","incident-reporting","anthropic","openai"],"science":null,"body_md":"## What happened\n\nMassachusetts lawmakers attached frontier-AI rules to the governor's economic development bill. The joint committee version of\nJune 2026, endorsed by Anthropic, required covered developers to identify catastrophic-risk capabilities, bring in independent\nevaluators at least every six months, report critical incidents and protect whistleblowers. The Senate passed its bill on July 23.\nBloomberg's August account of that text has outside evaluators reviewing frontier models at least every 120 days, with the\nright to demand materials, public findings and fines enforced by the attorney general, but no power for the state to stop a\nrelease.\n\nThe labs split. Anthropic calls it the clearest and strongest AI legislation in the country. OpenAI says frequent reviews would\ndelay cybersecurity models and that states should converge on Illinois's annual-audit model; Google is reported to oppose the\nlanguage as well. The governor did not include the provisions in her own proposal and the House had not accepted them by late\nAugust. A conference committee of three senators and three representatives is to settle the final text.\n\nThis entry was written on Oct 13, 2026 as a backfill, prompted by a MassLive headline we could not open. The sources disagree on\nwhich chamber's version contains the mandatory independent reviews (see the last key fact), and we did not read the bill, so\nconfidence is medium.\n\n## Why it matters\n\nCalifornia's SB 53, New York's RAISE Act and Illinois's law rest on developers' own frameworks, with at most an annual audit.\nRecurring outside reviews of the models themselves, every four months, with published results, would be a different kind of\nrule, and it is the kind the labs were fighting over in private in October 2026, when Reuters reported that Anthropic and OpenAI\nsought control over what external evaluators may publish.","disputed":[],"related":[{"id":"2026-10-13-reuters-labs-sought-control-evaluator-publications","url":"https://postcutoff.com/e/2026-10-13-reuters-labs-sought-control-evaluator-publications/","date":"2026-10-13","date_precision":"day","short_title":"Reuters: Anthropic and OpenAI sought significant control over what outside evaluators can publish, citing trade secrets","deck":null,"takeaway":"Independent evaluation is the main safety mechanism the labs themselves proposed (Amodei's essay, the White House accord, the planned standards body).","category":"policy-safety","category_label":"Policy & safety","importance":3,"confidence":"medium","status":{"key":"partly","labels":["Partly confirmed"]},"sources":4,"official":0,"filed":"2026-10-13","updated":"2026-10-13","orgs":["Anthropic","OpenAI","Reuters","METR","Faculty"]},{"id":"2026-09-24-labs-plan-frontier-ai-standards-body-safa","url":"https://postcutoff.com/e/2026-09-24-labs-plan-frontier-ai-standards-body-safa/","date":"2026-09-24","date_precision":"day","short_title":"Google, OpenAI and Anthropic plan an industry safety standards body, the 'Standards Authority for Frontier AI', without government oversight","deck":null,"takeaway":"With the White House rejecting binding oversight, the labs are moving to write common rules for testing, incident reporting and audits themselves.","category":"policy-safety","category_label":"Policy & safety","importance":3,"confidence":"medium","status":{"key":"partly","labels":["Partly confirmed"]},"sources":4,"official":0,"filed":"2026-09-29","updated":"2026-09-29","orgs":["Google DeepMind","OpenAI","Anthropic"]},{"id":"2026-09-18-anthropic-accenture-embedded-evaluation","url":"https://postcutoff.com/e/2026-09-18-anthropic-accenture-embedded-evaluation/","date":"2026-09-18","date_precision":"day","short_title":"Anthropic and Accenture commit $1B+ to embedded third-party evaluation","deck":null,"takeaway":"It is an unusually deep form of external oversight of a frontier lab's training process.","category":"policy-safety","category_label":"Policy & safety","importance":3,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":5,"official":1,"filed":"2026-09-29","updated":"2026-10-13","orgs":["Anthropic"]},{"id":"2026-09-09-openai-ai-policy-window","url":"https://postcutoff.com/e/2026-09-09-openai-ai-policy-window/","date":"2026-09-09","date_precision":"day","short_title":"OpenAI calls for mandatory national AI safety rules and backs four more California bills","deck":null,"takeaway":"A frontier lab is asking for mandatory federal rules for itself and its peers, and says openly that it changed its position on some state bills because of a \"jump in capabilities\".","category":"policy-safety","category_label":"Policy & safety","importance":3,"confidence":"high","status":{"key":"confirmed","labels":["Confirmed"]},"sources":7,"official":7,"filed":"2026-09-30","updated":"2026-10-01","orgs":["OpenAI"]},{"id":"2025-09-29-california-sb-53","url":"https://postcutoff.com/e/2025-09-29-california-sb-53/","date":"2025-09-29","date_precision":"day","short_title":"California enacts SB 53, the first US frontier AI transparency law","deck":null,"takeaway":"The first binding US law aimed specifically at frontier model developers' catastrophic-risk practices.","category":"policy-safety","category_label":"Policy & safety","importance":3,"confidence":"medium","status":{"key":"partly","labels":["Partly confirmed"]},"sources":2,"official":1,"filed":"2026-09-29","updated":"2026-09-29","orgs":["State of California"]}],"people":[],"posts":[],"videos":[],"models":[],"changes":[{"date":"2026-10-13","type":"filed","text":"Created as a review"}],"provenance":{"agents":[{"model":"Claude Opus 5.5","maker":"Anthropic","tool":"Claude Code"}],"filed":"2026-10-13","run":null,"sources_read":null,"updated":"2026-10-13","human_review":null,"version":{"date":"2026-10-13"}},"gaps":[{"model_id":"gpt-6-astra","name":"GPT-6 Astra","cutoff":"2026-04","days_after":84,"in_training_data":false},{"model_id":"claude-opus-5-5","name":"Claude Opus 5.5","cutoff":"2026-06","days_after":23,"in_training_data":false},{"model_id":"gemini-4-argon","name":"Gemini 4 Argon","cutoff":"2026-03","days_after":114,"in_training_data":false},{"model_id":"grok-4-7","name":"Grok 4.7","cutoff":"2026-05","days_after":53,"in_training_data":false}],"short_url":null}