⚖️ Legal Landscape

Tracking bans, regulations, and lawsuits around AI use globally — how policy shapes adoption and resistance across jurisdictions.

Sources last verified

Regulation in Motion

As AI capabilities expand, governments worldwide are grappling with how to regulate these systems. The legal landscape is evolving rapidly, with different jurisdictions taking varied approaches that shape how AI develops and deploys.

European Union

The AI Act

The EU AI Act (Regulation 2024/1689), which entered into force on August 1, 2024, represents the world's first comprehensive AI regulation. Key provisions include:

  • Risk-based classification — Four tiers: unacceptable, high, limited, and minimal risk
  • Prohibited applications (effective February 2025) — Bans on social scoring, untargeted facial recognition scraping, emotion recognition in workplaces/schools
  • Transparency requirements (Article 50, effective August 2, 2026) — Users must know when interacting with AI chatbots or deepfakes
  • High-risk oversight — Strict requirements for AI in healthcare, employment, law enforcement and biometrics. These deadlines moved in 2026; see below.

Real-time biometric identification is banned except for narrow law enforcement exceptions (missing persons, imminent threats, terrorist attacks).

The 2026 Digital Omnibus Deferral

The high-risk timetable originally set for August 2, 2026 no longer applies. The European Parliament endorsed a deferral package on June 16, 2026, the Council gave final approval on June 29, 2026, and the resulting "Digital Omnibus" was published in the Official Journal on July 24, 2026, entering into force on July 27, 2026 — days before the deadline it postponed.

  • Standalone high-risk systems (Annex III) — compliance deadline moved from August 2, 2026 to December 2, 2027
  • AI embedded in regulated products (Annex I) — moved to August 2, 2028
  • Article 50 transparency duties — not deferred; these still applied from August 2, 2026

The distinction matters: the chatbot-disclosure and deepfake-labelling obligations that most people encounter took effect on schedule, while the conformity-assessment regime for high-risk deployment slipped by 16 months. (Sources: Gibson Dunn; Jones Walker)

United States

Sectoral Approach

The US has eschewed comprehensive AI legislation in favor of sector-specific rules and executive guidance. Key developments include:

  • Executive Order 14110 (October 30, 2023) — Biden's comprehensive AI safety order was revoked by executive order on January 20, 2025, leaving federal AI policy in flux
  • NIST AI Risk Management Framework — Voluntary guidance for identifying and managing AI risks (remains in effect)
  • FTC Operation AI Comply (September 25, 2024) — Enforcement sweep against deceptive AI claims under existing consumer protection law, opening with actions against five companies
  • State-level action — California Governor Newsom signed 18 AI bills into law in September 2024 — covering training-data disclosure (AB 2013, which took effect January 1, 2026), AI transparency and watermarking (SB 942, effective August 2, 2026), digital replicas, AI-voice robocalls and deepfakes — while vetoing SB 1047 and later signing the narrower frontier-model safety disclosure law SB 53 (effective January 1, 2026). Illinois's AI employment-discrimination law (HB 3773) took effect on January 1, 2026, while Colorado delayed its 2024 AI Act (SB 24-205) to June 30, 2026 before repealing and replacing it in May 2026 with a narrower Automated Decision-Making Technology disclosure statute (SB 26-189, effective January 1, 2027)

The lack of comprehensive federal legislation means states are leading on AI regulation, creating a patchwork of requirements across jurisdictions.

China

Centralized Control

China has implemented AI regulations focused on content control and social stability, including:

These regulations prioritize state interests over individual rights, reflecting China's broader approach to technology governance.

Key Legal Battles

Copyright and Training Data

Dozens of lawsuits challenge whether AI training on copyrighted material constitutes infringement; the running case tracker maintained by Edward Lee is the most complete public tally. Key cases include:

  • Authors Guild v. OpenAI — Class action alleging mass ingestion of literary works to train ChatGPT
  • The New York Times v. OpenAI & Microsoft — Alleges millions of articles were used without consent; seeks billions in damages
  • Andersen v. Stability AI — Visual artists challenging Stable Diffusion and Midjourney
  • Getty Images v. Stability AI — Claims infringement of 12+ million photographs
  • RIAA v. Suno (June 2024) — First major lawsuit over AI music generation

The June 2025 Fair Use Rulings

Two summary judgment decisions issued two days apart in the Northern District of California became the first substantive judicial treatment of whether LLM training is fair use — and they split on the reasoning that matters most.

  • Bartz v. Anthropic (Judge Alsup, June 23, 2025) — training on lawfully acquired books was "quintessentially transformative" and fair use, but building and retaining a permanent library of pirated copies was a separate, non-transformative use that fair use did not protect (analysis)
  • Kadrey v. Meta (Judge Chhabria, June 25, 2025) — training held "highly transformative" and fair use, with the court declining to make the outcome turn on whether the underlying copies came from shadow libraries (analysis)

It was the surviving piracy holding in Bartz — not the training question — that set up a damages trial and led Anthropic to settle in August 2025, the first major settlement in this wave of cases. The pattern to watch is that acquisition practices, rather than training itself, are so far where liability has actually attached.

Employment Discrimination

Landmark cases are establishing liability for AI-based hiring discrimination:

  • EEOC v. iTutorGroup (2023) — First AI discrimination settlement ($365,000); software automatically rejected applicants based on age
  • Mobley v. Workday — Collective action alleging AI screening discriminates by race, age, and disability; the court granted preliminary certification of a nationwide age-discrimination collective in May 2025 and denied Workday's motion to dismiss the Third Amended Complaint in June 2026, allowing nationwide disparate-impact claims to proceed into discovery
  • July 2024 ruling — In the same case, the court held that an AI vendor can be liable as an agent of the employers using its tool, not merely the employers themselves

The EEOC has made AI bias an enforcement priority. On the research side, a University of Washington audit by Kyra Wilson and Aylin Caliskan tested three resume-screening models across 554 resumes and 571 job descriptions — more than three million comparisons — and found white-associated names preferred 85% of the time against 8.6% for Black-associated names, with Black-male names never preferred over white-male names. The paper was presented at the AAAI/ACM Conference on AI, Ethics, and Society in October 2024 (University of Washington; Wilson & Caliskan, arXiv:2407.20371).

Liability and Accountability

When AI systems cause harm, who is responsible? Courts are addressing:

  • Autonomous vehicle accidents
  • Medical AI misdiagnoses
  • Financial AI causing market disruption

Current legal frameworks struggle to assign responsibility for decisions made by AI systems.

Workplace Protections

Union Responses

Labor organizations are negotiating AI-related protections:

  • Requirements for advance notice of AI implementation
  • Worker input on AI tool selection
  • Protections against AI-based surveillance and management
  • Retraining guarantees for affected workers

Emerging Legislation

Some jurisdictions are considering or implementing:

  • Right to human review of AI decisions
  • Algorithmic transparency requirements
  • AI impact assessments before deployment
  • Worker consent for AI monitoring

Looking Ahead

The legal landscape for AI remains unsettled. Key questions include:

  • Will international standards emerge, or will fragmentation continue?
  • How will courts adapt existing legal concepts to AI?
  • Can regulation keep pace with technological change?
  • Will worker protections be strengthened or weakened?

The answers will significantly shape how AI impacts society, making legal developments essential to understanding technology resistance.

References