Law and AI Briefing — Stone Arch Collective

The Stone Arch Bridge spanning the Mississippi River in Minneapolis — Stone Arch Bridge, Minneapolis — via Wikimedia Commons (placeholder — swap for one of ours).

Today's Legal Signal

Connecticut's AI Responsibility and Transparency (CART) Act's frontier-developer whistleblower protections, automated-employment-decision-technology (AEDT) disclosure duties, and AI-subscription/synthetic-content transparency rules all go live today, October 1 — additional parts of the law take effect, including disclosure requirements around AI subscriptions, new whistleblower protections for people involved with the development and deployment of large AI frontier models, and new definitions standardizing AI technology in the state, giving employers and AI vendors their first hard compliance deadline since the Third Circuit's sealed fair-use ruling landed this week.

Today

AI & Law News

01
Connecticut's CART Act provisions take effect today: covered employers using automated employment-decision technology must give plain-language notice, and large frontier AI developers must stand up anonymous whistleblower-reporting channels for 'catastrophic risk' outputs. Frontier development employees must be allowed to anonymously report any AI development or outputs that might pose a 'catastrophic risk' to the public, defined as materially contributing to the death of, or serious injury to, more than 50 people, or more than $1 billion in damage. CT Public / EBG Law
02
The Connecticut Supreme Court has moved from warnings to formal process: it has threatened to sanction GLG Law LLC and one of its attorneys for submitting documents in two cases that misrepresented the law through the use of generative artificial intelligence, summoning them to appear in court next month. Law360 Pulse
03
A federal judge pushed back on the assumption that AI research tools are a substitute for lawyer judgment: a Connecticut federal judge urged attorneys during a sanctions hearing to push back against clients who demand lawyers use generative AI tools for legal research, saying the technology 'aims to please' and can misstate the law. Law360 Pulse
04
Legal-tech company Legion has escalated a dispute over export-control-style AI restrictions into federal court: Legion sued the U.S. government in D.C. federal court over a directive ordering Anthropic to shut down two of its advanced AI models to foreigners, alleging the move caused the company to lose access to a model powering its platform. Law360 Pulse
05
Legal-ops consolidation continues as AI reshapes the ALSP (alternative legal service provider) market: Repario, a company focused on e-discovery and managed review, has acquired UnitedLex, a global provider spanning legal process outsourcing, litigation, IP, contracts and technology. LawSites
06
A new proposed class action alleges Anthropic, OpenAI, xAI and Google unlawfully coordinated to slow frontier model development: the suit, filed in the U.S. District Court for the Northern District of California, argues the companies violated antitrust law by agreeing to coordinate slowdown efforts, reducing the value consumers get for paid AI subscriptions. Associated Press via CNN

Past 7 days

September 30, 2026

01
The Third Circuit affirmed the Delaware district court's rejection of ROSS Intelligence's fair-use defense in its long-running Westlaw headnotes dispute with Thomson Reuters; the Third Circuit on Tuesday affirmed Thomson Reuters' landmark copyright win against Ross Intelligence, leaving intact a Delaware federal court ruling that the now-defunct legal research startup infringed by training its AI search tool on Westlaw material. The panel's reasoning is not yet public: the 3rd U.S. Circuit Court of Appeals today issued an opinion in the long-running copyright battle between Thomson Reuters and the now-shuttered AI legal research company ROSS Intelligence, but the opinion is, at least for now, sealed and not viewable by the public. Law360 / Legal IT Insider
02
A new antitrust theory hit the AI industry this week: the lawsuit, which was filed Friday in the U.S. District Court for the Northern District of California, argues that the leading AI companies violated antitrust laws when they agreed to coordinate slowdown efforts, and that doing so would reduce the value consumers get for paid AI subscriptions. Lawyers representing four named plaintiffs, who pay for subscriptions to ChatGPT, Claude, Grok or Gemini, are bringing the lawsuit on behalf of a proposed nationwide class of other paid subscribers to those services. Associated Press (via CNN)
03
Product-liability exposure for consumer chatbots is scaling fast: on April 29, 2026, Edelson PC filed seven lawsuits on behalf of its clients in the United States District Court for the Northern District of California against OpenAI and CEO Sam Altman, and in September 2026, an additional 30 lawsuits, bringing the total to 37 cases against OpenAI. Edelson PC
04
AI-washing securities litigation keeps climbing: in recent days, plaintiffs' lawyers have filed two further AI-related securities class action lawsuits, first, against the Chinese Internet company Baidu, and, separately, against the AI-powered Internet advertising firm AppLovin. On September 16, 2026, a plaintiff shareholder filed a securities class action lawsuit in the Northern District of California against AppLovin and certain of its directors and officers. By our tally, these new lawsuits bring the 2026 count of AI-related lawsuit filings to 24, representing nearly 14% of all new securities class action lawsuit filings this year. D&O Diary
05
On the international legal-tech front, a jurisdiction-specific legal AI assistant launched in the UK this week: Alesis, an AI assistant grounded solely in the laws of England and Wales, Scotland and Northern Ireland, and which checks its answers "exhaustively" against primary sources, has gone live, a notable entrant as firms weigh hallucination and sourcing risk in jurisdiction-bound legal AI tools. Legal IT Insider
06
The 'new law firm' model keeps gaining scale: Pierson Ferdinand, a new model law firm that has no associates and instead relies of AI platforms such as Harvey to do the heavy lifting, reportedly reached 300 partners this week — a data point for GCs and firm leaders assessing AI-driven staffing and pricing models. Artificial Lawyer

Regional

Global Watch

01
EU enforcement has moved from theory to practice: as of August 2, 2026, the European Commission, through its European AI Office, holds formal investigative and enforcement authority over general-purpose AI model providers and the Act's prohibited practices, and has already used it. The AI Office opened its first round of formal investigations in June 2026, targeting AI systems deployed across European markets in hiring tools, credit scoring, and student-monitoring applications. Volkov Law Group
02
A Canadian provincial government has taken the novel step of suing an AI developer directly over a mass-casualty event: on September 21, the British Columbia government sued OpenAI over the 2026 Tumbler Ridge shooting that killed eight, alleging the shooter used ChatGPT to plan the attack. Wikipedia (2026 in artificial intelligence), citing contemporaneous news reporting

Past 7 days

September 30, 2026

01
The EU's AI Office moved from theoretical to active enforcement this month: after the AI Office's enforcement powers with regard to general-purpose AI models under the AI Act entered into application on 2 August, the Commission followed promptly with a first round of requests for information to different AI companies focusing both on safety and security aspects as well as copyright-related obligations. Center for Democracy & Technology Europe
02
Transparency-rule implementation is also moving: transparency obligations for providers and deployers of publicly-facing AI tools are similarly applicable since 2 August, and requirements to comply with the AI Act in this context were further detailed in a code of practice which was signed by 190 organisations, including major providers. The European Commission also published an AI Act complaints tool which enables individuals and organisations to submit complaints to the AI Office about alleged infringements of the AI Act, giving EU counsel a new procedural mechanism to flag client or competitor non-compliance. Center for Democracy & Technology Europe
03
Separately, on the platform-regulation side, the European Commission announced that it had designated ChatGPT as a Very Large Online Search Engine (VLOSE), and Reddit and Roblox as Very Large Online Platforms (VLOPs), under the Digital Services Act. VLOSE status brings enhanced obligations, including assessing and mitigating systemic risks arising from the service and its algorithmic systems, covering areas such as the spread of illegal content, effects on minors, users' physical and mental wellbeing, fundamental rights, electoral processes and public security, with compliance due by January 2027 — a cross-regime overlay (DSA plus AI Act) that EU-facing GPAI providers now have to map simultaneously. European Commission

Practice

Try This in Your Practice

Map your CART Act exposure in 10 minutes

The 10-minute versionPull up the CART Act's October 1 provisions and check three things against your organization's or client's practices: (1) does any AI-subscription product give proper renewal notice, (2) does any AEDT used in hiring/firing come with the required plain-language disclosure, and (3) does your org meet the $500M revenue threshold that triggers frontier-developer whistleblower obligations.

Level upBuild a one-page CART Act compliance checklist mapping each operative section (whistleblower, AEDT, synthetic content, subscriptions) against your AI vendor contracts and flag gaps for GC review before Q4 close.

Learn more — Read the CT Public breakdown of what takes effect Oct. 1

Citation-check your last AI-assisted filing

The 10-minute versionTake the most recent AI-drafted memo, brief, or client letter in your files and run every citation through a manual or Westlaw/Lexis verification pass, looking specifically for hallucinated case names or misstated holdings.

Level upDraft or update a firm-wide AI-use policy requiring a documented, signed-off verification step before any AI-assisted filing goes out the door, citing the Connecticut Supreme Court's GLG Law sanctions order as the cautionary precedent.

Learn more — See the Law360 coverage of the GLG Law sanctions order

Past 7 days

September 30, 2026

Run a 10-minute AI fair-use exposure scan

The 10-minute versionList every AI tool (internal or vendor) your firm or client uses that was trained on a proprietary or licensed dataset (case law compilations, industry databases, internal knowledge bases). For each, ask: is the training use transformative, or does it substitute for the original compiled product — the exact line the Third Circuit just drew against ROSS Intelligence?

Level upPull the vendor's data-training disclosures or DPA/MSA language and check for a copyright indemnification clause; if none exists, flag it to procurement before renewal.

Learn more — Read the Third Circuit coverage

Spot-check EU high-risk AI exposure in HR/recruiting tools

The 10-minute versionEmail your (or your client's) HR/recruiting lead one question: does any AI tool used to screen, rank, or score candidates process applications from people located in the EU? If yes, that tool is presumptively high-risk under the AI Act's Annex III and needs a documentation review.

Level upRequest the vendor's conformity assessment, technical documentation, and human-oversight protocol, then compare against the AI Office's published GPAI/high-risk guidance before the next contract renewal cycle.

Learn more — See the EU AI Office's September bulletin

Implications

What This Means for Practitioners

01Connecticut's CART Act provisions effective today create the first live, operative frontier-AI whistleblower channel and AEDT disclosure regime in the US — in-house counsel at any large AI developer or employer using automated hiring/firing tools with Connecticut touchpoints should confirm compliance now, not at year-end.
02Courts are converting warnings about AI hallucinations into real sanctions exposure (Connecticut Supreme Court vs. GLG Law) — firms without a documented AI-citation-verification protocol are increasingly exposed to Rule 11-style sanctions, not just reputational embarrassment.
03The EU AI Office's active investigations into hiring, credit-scoring, and student-monitoring AI mean multinational clients deploying these systems in Europe face real enforcement risk now, regardless of where the company is headquartered — this should be on every GC's EU-market AI risk register.
04A government directly suing an AI developer over a mass-casualty event (British Columbia v. OpenAI) signals that product-liability and failure-to-warn theories against chatbot makers are moving from academic discussion to live litigation with state/provincial plaintiffs, raising the stakes for any company offering consumer-facing conversational AI.

Watch

Case & Enforcement Watch

Thomson Reuters v. ROSS Intelligence3rd Cir., No. 25-2153 — opinion issued Sept. 29, 2026 (sealed)
Third Circuit affirmed Thomson Reuters' fair-use win in the first federal appellate ruling on AI training and fair use; opinion issued September 29 but remains sealed pending redaction review.
LawSites
In re: AI Slowdown Antitrust LitigationN.D. Cal. — filed Sept. 19, 2026
New proposed class action alleges Anthropic, OpenAI, xAI (formerly SpaceXAI) and Google illegally coordinated to slow frontier AI development, harming paid-subscription consumers.
Associated Press via CNN
British Columbia v. OpenAICanada — filed Sept. 21, 2026
BC government sues OpenAI alleging the 2026 Tumbler Ridge mass shooter used ChatGPT to plan the attack that killed eight.
Wikipedia (2026 in artificial intelligence)
New York Times v. Microsoft & OpenAISDNY — DOJ brief filed Sept. 1, 2026
DOJ filed a statement of interest backing OpenAI's fair-use defense — first time the federal government has taken a position in AI-training copyright litigation; case remains in active MDL proceedings in SDNY.
Reuters / Wikipedia
Concord Music Group v. AnthropicN.D. Cal. — cross-MSJ briefing as of Aug. 24, 2026
Cross-motions for summary judgment on fair use now fully briefed in the music publishers' lyrics-training case against Anthropic.
McKool Smith AI Litigation Tracker
Getty Images v. Stability AIUK High Court — judgment Nov. 4, 2025
UK High Court rejected Getty's core copyright claim (model weights not an infringing 'copy') but found limited trademark infringement from watermark reproduction — the leading UK precedent on generative AI and copyright.
regulations.ai UK AI Regulation Overview
ICO v. Clearview AIUK Upper Tribunal — Oct. 6, 2025
UK Upper Tribunal restored the ICO's jurisdiction to enforce its £7.6M fine against Clearview AI for unlawfully scraping facial images into its recognition database.
regulations.ai UK AI Regulation Overview

Weekly — updates Mondays

Global Regulatory Watch

Nothing here yet — this section updates every Monday.

A note on accuracy: this briefing is AI-assisted research. Double-check anything you'd cite, quote, or act on — especially numbers, dates, and direct claims.