European Commission published draft guidelines on high-risk AI classification covering employment, education, critical infrastructure, and law enforcement—with 55 days to enforcement deadline. Misclassification now carries regulatory intervention and penalty risk; builders deploying in EU need immediate compliance audit of intended purpose and real-world deployment context.

New Mexico jury held Meta liable for insufficient safeguards against sexual predators; California jury found intentional addictive design; now AI companies face similar "Big Tobacco" litigation pattern with no federal standard. Absence of uniform safety rules is driving state-level and private lawsuits that could establish precedent for platform liability across AI products.

Disney Enterprises and Universal filed suit against Midjourney in June 2025 alleging systematic infringement—training on copyrighted characters and images without authorization. Training data provenance is now actively litigated; model builders face discovery risk on scraping practices and need explicit licensing or fair-use defense strategy.

AFM lawsuit accuses Warner and Universal of licensing human talent to AI training platforms (Suno, Udio) without sharing settlement proceeds or future revenue with source artists. Copyright holders are now directly suing labels and platforms for revenue split; this sets precedent for artist compensation in AI training economics.

OpenAI asked federal judge to dismiss lawsuit alleging it gave unauthorized legal advice, arguing ChatGPT is not practicing law. Product liability and unauthorized practice claims are escalating; platform teams need clear terms-of-service boundaries on professional advice and legal disclaimers to reduce regulatory exposure.