AI Software

China Deep Synthesis Provisions

CAC Deep Synthesis Provisions require labelling of synthetic media, algorithm registration, and real-name registration of users. Together with the 2022 Algorithm Recommendation Provisions, sets the architecture for China's AI model registration regime that pre-dates the EU AI Act.

CAC Generative AI Measures take effect

Cyberspace Administration of China (CAC) issues Interim Measures for the Management of Generative AI Services. Public-facing GenAI services must complete CAC algorithm registration, security assessment, and content moderation aligned with socialist core values. Foundational to subsequent CAC enforcement against domestic LLM providers.

EU AI Act enters into force

The EU AI Act enters into force after publication in the Official Journal on 12 July 2024. The Act introduces a risk-tiered framework (prohibited, high-risk, limited-risk, minimal-risk) and a separate regime for general-purpose AI (GPAI) models. Most obligations phase in over 6 to 36 months.

EU AI Act prohibitions take effect

Article 5 prohibitions become applicable: social scoring, manipulative AI, exploitation of vulnerabilities, untargeted facial-recognition scraping, certain real-time remote biometric identification in public, emotion recognition at work or school, and predictive policing based solely on profiling.

EU AI Act GPAI obligations apply

General-purpose AI model obligations apply: technical documentation, training-data summaries, copyright policy, and systemic-risk obligations for the largest models (FLOPs threshold of 10 to the 25th). The Commission's AI Office and the GPAI Code of Practice are the primary instruments.

EU AI Office becomes operational

The EU AI Office (inside DG Connect) coordinates enforcement of the AI Act's GPAI provisions, the GPAI Code of Practice, and the AI Pact voluntary regime. Performs model evaluations and oversees the systemic-risk GPAI category.

EU AI Act high-risk obligations apply

Most high-risk AI system obligations become enforceable: conformity assessments, risk management systems, data governance, human oversight, post-market monitoring, and registration in the EU database. Covers credit scoring, hiring, education, critical infrastructure, law enforcement, and biometrics.

EU AI Act embedded high-risk obligations apply

Final phase: high-risk obligations extend to AI systems embedded as safety components in products already regulated under EU product-safety law (medical devices, machinery, toys, vehicles). Completes the multi-year AI Act compliance ramp.

UK AI Safety Institute established

Established at the Bletchley Park AI Safety Summit, the UK AI Safety Institute (AISI) became the first government-backed body to evaluate frontier model capabilities and safety. AISI has signed pre-deployment evaluation agreements with OpenAI, Anthropic, Google DeepMind, Microsoft, and Meta. Renamed AI Security Institute in early 2025.

Authors Guild v. OpenAI filed

The Authors Guild and 17 named authors (including George R.R. Martin, John Grisham, Jodi Picoult) file a class action against OpenAI in S.D.N.Y. for use of copyrighted books in training data. Consolidated with related class actions in 2024.

NYT v. OpenAI and Microsoft filed

The New York Times sues OpenAI and Microsoft in S.D.N.Y., alleging mass copyright infringement in training data and output regurgitation. Most-watched AI copyright case; OpenAI's motion to dismiss was largely denied in 2024, with discovery proceeding into 2025 and 2026.

US AI Safety Institute established

NIST established the US AI Safety Institute (US AISI) inside NIST, with voluntary pre-deployment testing agreements signed with OpenAI and Anthropic in August 2024. Mandate under review following the rescission of EO 14110 and the 2025 AI Action Plan.

RIAA v. Suno and Udio filed

The Recording Industry Association of America (Sony Music, UMG, Warner) sues Suno (D. Mass.) and Udio (S.D.N.Y.) alleging mass training-data infringement of sound recordings. First major music-industry test of generative AI training-data liability.

BIS AI Diffusion Framework published

The Bureau of Industry and Security (BIS) publishes the Interim Final Rule on AI Diffusion: tiered global access to advanced AI compute and model weights via a country-cap and Validated End User regime. Builds on October 2022 and October 2023 export-control rules targeting H100, H200, B200 sales to China. Subsequently paused for further review under the Trump administration.

Trump signs EO 14179, rescinds Biden AI EO 14110

Executive Order 14179 (Removing Barriers to American Leadership in Artificial Intelligence) rescinds Biden's EO 14110 (Oct 2023) and directs OSTP and OMB to draft a new AI Action Plan within 180 days. Centerpiece: deregulatory posture, reduced reporting requirements, focus on US competitive position versus China.

White House releases AI Action Plan

The OSTP and OMB release the AI Action Plan called for by EO 14179. Plan emphasizes federal-procurement preference for American AI, accelerated permitting for data centers, reform of NIST AI Safety Institute mandate, and limited federal preemption of state AI laws (notably Colorado AI Act).

Colorado AI Act takes effect

Colorado SB 24-205 becomes the first US comprehensive AI consumer-protection law to take effect. Imposes duty of reasonable care on developers and deployers of high-risk AI systems (used in consequential decisions: hiring, lending, housing, education, healthcare, insurance, legal services). Requires impact assessments and consumer notice and appeal rights.

Getty Images v. Stability AI

Getty Images files parallel suits in the High Court of Justice (UK) and the District of Delaware against Stability AI over Stable Diffusion training data. UK trial proceeded in mid-2025; US Delaware case advanced into discovery. Landmark test of image-training copyright liability.