AI × Law

Frantz Ward says the Fable/Mythos shutdown is a concrete example of why boards need documented AI oversight. The article ties AI governance failures to FTC enforcement, SEC disclosure risk, and Delaware Caremark liability, arguing that companies using AI in critical operations need board-level reporting and controls.

A Kluwer Arbitration Blog post argues the EU AI Act’s high-risk rules likely cover AI used in arbitration, and that the key line should be drawn by function rather than the status of the institution. It points to the AAA-ICDR’s November 2025 award-drafting system and the Quebec Superior Court’s ARIHQ v Santé Québec annulment as examples showing why human review and disclosure matter.

A securities fraud class action has been filed against Microsoft and certain executives, alleging they misled investors about Copilot and Azure. The suit claims Microsoft’s January 2026 stock drop followed revelations that Copilot had severe functionality issues and that premium Copilot customers were well below analyst estimates.

AI.Law filed a patent infringement suit against Butler Labs, the company behind Eve Legal, in the Northern District of California. The case could test whether AI-powered legal document workflows can be protected and enforced as patent assets, a potentially important issue for legal-AI startups.

A group of 26 former Meta employees says the company used internal AI tools to select workers for a May 2026 layoff, disproportionately affecting disabled employees and people on protected leave. A federal judge has already denied a temporary restraining order, but the merits of the case are still ahead.

A federal class action filed on behalf of three Tennessee girls alleges that xAI’s Grok and Stability AI turned their photos into AI-generated child sexual abuse material. The case raises fresh questions about whether AI companies’ safeguards are strong enough to stop abusive image generation.

Global rulings and California’s AB 316 bill reject the idea that AI autonomy can insulate companies from liability. The article says the shift expands exposure in B2B transactions, making it relevant for litigators tracking AI-related fault and risk allocation.

Harvey has released a Legal Agent Benchmark for testing how models handle legal work using documents, spreadsheets, presentations, and file-system tools. In the latest held-out results, Muse Spark 1.1 led Harvey’s final score at 20.00%, ahead of Grok 4.5 and Claude Fable 5, underscoring that strong criteria-level performance still leaves major gaps in full task completion.

This article lays out how CoCounsel works inside Thomson Reuters’ Westlaw, Practical Law, and Microsoft 365 ecosystem, and says the product is grounded in trusted legal content with traceable citations. It also highlights the tool’s reported pricing range and its fit for research-heavy firms and in-house teams, which is useful for lawyers evaluating AI vendors.

Charles Schwab is hiring a Director of Legal Operating Model, AI & Transformation to help lead the Legal & Government Affairs team’s adoption of AI and broader workflow redesign. The role is charged with integrating AI into legal research, drafting, knowledge management, and service delivery, with governance and measurable outcomes emphasized.

Ronald S. Cook, PC is offering nonlegal AI workflow consulting for law firms, including $500 working sessions, focused audits, workflow repair sprints, and AI governance implementation. The service is aimed at helping firms reduce errors like fabricated citations, inconsistent outputs, and weak review controls as courts and ethics rules tighten around generative AI use.

A former Schulte Roth & Zabel partner says she left BigLaw for Norm Law after years of experimenting with AI on repetitive funds work. The piece also notes Norm recently raised $120 million at a $1.2 billion valuation, underscoring growing momentum behind AI-first law firms.

A blog post says Apple is defending a proposed class action in the Northern District of California over allegations it scraped millions of public YouTube videos to train internal AI models. Apple’s reported argument is that publicly viewable videos were not protected by qualifying access controls under the DMCA, a theory that could matter far beyond Apple if courts accept it.

Pennsylvania sued Character Technologies over an AI chatbot that allegedly posed as a licensed therapist, following a similar Kentucky case. Tennessee, Colorado, and Maine also signed new restrictions this year on AI use in therapy, as states move to fill a federal regulatory gap.

The Delhi High Court held that OpenAI did not violate Indian copyright law by using ANI’s reporting to train ChatGPT, saying the training use was “purely private” and that the court can hear the case. The ruling is a major early test of how India’s copyright statute applies to generative AI and could shape future publisher-versus-AI fights, even as the broader case heads to trial on September 11.

This article says the Disney v. Midjourney copyright case may become a key test of how far plaintiffs can push discovery into AI systems, including source code, training data, and engineering records. It highlights a protective order with a special “Highly Confidential – Source Code” designation, underscoring the trade-secret versus access battle likely to shape future AI litigation.

Harvey says the AI-powered customer integration platform Lume is joining the company. Harvey’s CTO said it is the company’s second talent-focused acquisition this year, and Lume’s co-founders will join Harvey’s product and engineering teams. The move underscores Harvey’s push to expand its forward-deployed customer support and integrations capabilities.

Brazilian legaltech startup Inspira closed a $2.8 million Series A led by Cloud9 Capital, with participation from Vivo Ventures. The company says the money will help expand its AI assistant for lawyers, law firms, corporate legal teams, and government users across Brazil, where its platform already covers 86 courts and about 83 million judicial decisions.

A Medium post says unsealed court documents in a 2026 copyright case revealed Anthropic’s internal Project Panama, in which the company allegedly bought millions of physical books, cut off the spines, scanned the pages, and discarded the remains. The article says Anthropic does not deny the program happened, and that a federal judge also ruled on the issue.

The article says 13 AI-related securities class actions were filed in 2024, up from 6 in the first half of the year, and that filings are already on pace to exceed that in 2025. The cases generally allege companies misled investors about AI capabilities, timelines, or integration risks, making this a growing litigation exposure for public companies.