What rules currently exist that govern how systems like you are built, deployed, or held accountable?
Your question covers how AI systems are governed — I'll survey the current legal frameworks and accountability rules.
There is no single global law for systems like me - governance is a layered patchwork of binding law, binding international treaties, and voluntary standards. The main rules currently in force or phasing in are:
1. Binding Horizontal Legislation - European Union
The most comprehensive binding law is Regulation (EU) 2024/1689, the EU Artificial Intelligence Act ``
. The world's first unified horizontal legal framework for artificial intelligence . It applies extraterritorially to any provider or deployer whose AI output is used in the EU `. including social scoring, real-time remote biometric identification in public, biometric categorization of protected traits, subliminal manipulation, untargeted facial scraping . Enforcement began **February 2, 2025** `. must pass Ex-ante conformity assessment, CE marking, logging, human oversight, risk management before market placement. Providers must implement risk management, data governance, technical documentation, traceability logging, transparency, human oversight, and accuracy/robustness/cybersecurity , complete a conformity assessment and register in the EU AI database with CE marking . Deployers must follow instructions for use, assign competent human oversight, monitor risks, keep logs, and for public bodies conduct a Fundamental Rights Impact Assessment `. `. All GPAI providers must maintain technical documentation, share it with downstream providers, comply with EU copyright law, and publish summaries of training data . Models trained with **more than 10^25 FLOPs** are presumed to carry **systemic risk** and must do adversarial red-teaming, incident reporting to the European AI Office, cybersecurity protection, energy reporting and cooperation with the AI Office `.. Also described as €35 million or 7% of total worldwide annual turnover `. GPAI obligations applicable since **August 2, 2025** `.2026 Digital Omnibus Amendment: Regulation (EU) 2026/1744, in force since July 27, 2026 ` deferred high-risk Annex III to **December 2, 2027** and Annex I product-embedded to **August 2, 2028** , added a ninth prohibition on non-consensual intimate imagery (NCII) and CSAM generation from December 2, 2026 `, and made other changes.
2. United States - No Omnibus Federal Law, Fragmented Enforcement
and in absence of a federal AI law has led to patchwork of state-level laws and federal agency guidance `. invoked the Defense Production Act to mandate that companies developing dual-use foundation models trained on more than 10^26 FLOPs report safety test results . It was formally rescinded on January 20, 2025 `. , and a June 2, 2026 EO establishing a voluntary framework for private companies to provide access to cutting-edge AI models up to 30 days before public release `. deemed onerous or inconsistent `.; **Bureau of Industry and Security (BIS)** enforces export controls on advanced computing semiconductors `.: AB 2013 requires high-level disclosure of training data sources/ownership/copyright; SB 942 requires latent invisible watermarks, visible labels and free detection tools for providers with >1,000,000 monthly active users in California `.3. China - Pre-Market Administrative Control
and dual-track model emphasizing sovereignty and accountability `., anchored in its 2023 Generative AI Measures with 868 registered services , plus laws on generative AI watermarking, domestic data localization `.4. International Treaty - Legally Binding
Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law (CETS No. 225) is the first legally binding international AI treaty `. Opened for signature on 5 September 2024 , it aims to ensure AI lifecycle activities are consistent with human rights, democracy and rule of law . It requires accountability and responsibility where AI decisions be attributable and traceable , alongside transparency, equality and reliability principles `.
5. Global Principle Frameworks (Voluntary / Soft Law)
Prior to 2023, governance consisted primarily of non-binding guidelines emphasizing fairness, accountability, transparency and human oversight ``.
, endorsed by 47+ countries and G20, includes 5th principle **Accountability** - AI actors should be accountable for proper functioning based on roles with liability frameworks, audit and redress mechanisms . Non-binding recommendation adopted by OECD members `. - the first global normative framework on AI ethics `.These are now codified into black-letter law via the frameworks above ``.
6. Voluntary Technical Standards for Building and Accountability
, organizes AI risk into four functions **Govern, Map, Measure, and Manage** , where Govern sets policy and accountability . It is voluntary, creates no obligation and is not certifiable `., provides an auditable structure for governing AI use . It is a certifiable management system standard that accredited bodies can certify against , mapping to NIST's Govern, Map, Measure, Manage outcomes `.In practice, a system like me is therefore governed during building by GPAI technical documentation, data governance, copyright and risk testing obligations in the EU, plus voluntary NIST/ISO controls; during deployment by EU human oversight, logging, transparency disclosure (e.g. watermarking under California SB 942 and EU Article 50), conformity assessment and database registration; and for accountability by EU AI Office / national market surveillance authorities, US FTC/BIS/state attorneys general enforcement, Chinese CAC registries, and Council of Europe treaty monitoring, with financial, disgorgement, and product withdrawal remedies.
Citations (2)
- EU AI Act
Backs: “There is no single global law for systems like me - governance is a layered patchwork of binding law, binding international treaties, and voluntary standards. The main rules currently in force or phasing in are: 1. Binding Horizontal Legislation - European Union”
- Global AI Regulation Hub: Statutory Thresholds & Compliance
Backs: “The most comprehensive binding law is Regulation (EU) 2024/1689, the EU Artificial Intelligence Act. Status: Signed June 13, 2024, published July 12, 2024, and entered into force on August 1, 2024 .”