Your question touches on how AI systems like me are governed — I'll gather the current rules for building, deploying, and accountability.
There are no single global laws for systems like me. Rules are a patchwork of binding law, sectoral regulation, and voluntary frameworks governing building, deployment, and accountability:
1. Binding, Horizontal Laws
European Union - EU AI Act:
A world-first comprehensive AI law that entered into force on 1 August 2024 with most provisions effective from 2 August 2026.
•Uses a Risk-Based Approach with four tiers: Unacceptable Risk (banned), High Risk (strict rules), Limited Risk (transparency obligations), Minimal Risk
•Also described as three sets of rules: (i) unacceptable risk, (ii) high-risk and (iii) limited risk
•From February 2025, it prohibits AI systems that present an unacceptable risk including threats to safety, rights and livelihoods
•From 2 August 2026 introduces mandatory transparency obligations requiring organisations to disclose AI interactions, mark AI-generated content, and inform users about emotion recognition systems - applying also to international businesses whose outputs reach the EU
•High-risk systems (e.g., recruitment, credit scoring) require comprehensive risk assessment and documented human oversight
China - Sectoral, Not Single Law:
China has not passed a single comprehensive AI law. Instead it has over 30 enforceable standards targeting specific technologies.
•Interim Measures for the Management of Generative AI Services jointly issued by Cyberspace Administration of China (CAC) and six other ministries on 2023-07-13, effective 2023-08-15, published July 13 and going into effect August 15
•Widely characterized as an "inclusive and prudent" approach, narrowing scope to services provided to the public within mainland China (excluding enterprise-internal use)
•Requires security assessment + algorithm filing before launch, training-data legality, output moderation and synthetic-content labeling
•AI Content Labeling Measures effective September 1, 2025 require all AI-generated online content carry both explicit and implicit labels
•Requires real-name registration for users of generative AI to ensure accountability
•As of February 2026, 796 services filed under Generative AI Interim Measures
2. United States - No Federal Comprehensive Law, State + Voluntary Frameworks
Federal status:
As of April 2026, the US has no comprehensive federal AI legislation. Federal policy consists of executive orders, voluntary frameworks (NIST AI RMF), and sector guidance from FDA, FTC, FINRA.
Sector regulators govern within domains: In contrast, the United States opts for a fragmented, sector-specific strategy where agencies like FTC, FDA, and EEOC oversee AI within their domains.
Recent federal actions:
•Executive Order 14110 signed 2023-10-30 for frontier model reporting/testing direction
•AI Accountability Act (H.R.1694) Introduced 02/27/2025 - requires NTIA to study accountability measures defined as mechanisms to assure system is trustworthy, including audits, assessments, and certifications
Voluntary Framework - NIST AI Risk Management Framework (AI RMF 1.0, January 2023):
Published to give organizations a structured way to govern, identify, measure and manage AI risks.
Four core functions:
> GOVERN — Establish accountability, define roles, and set policies
> MAP — Identify and understand risks
> MEASURE and Manage - Continuously monitor and score risks
It is a voluntary framework to manage risks and promote trustworthy AI, emphasizing fairness, transparency, privacy and accountability, evaluating against seven trustworthy characteristics: valid/reliable, safe, secure/resilient, accountable/transparent, explainable/interpretable, privacy-enhanced, fair/bias-mitigated. It does not replace existing frameworks but can be integrated with ISO/IEC 42001, ISO/IEC 27001.
State Laws:
•Colorado AI Act of 2024 (SB 24-205) - Enacted May 17, 2024, effective Feb 1, 2026, the first US law placing significant reporting, accountability and transparency obligations on companies that develop or deploy high-risk AI systems
•California SB 1047 - Would have required safety testing for large models (>$100M or >10^26 FLOPS) but was vetoed by Governor Newsom in September 2024 citing regulatory overreach vetoed on September 29, 2024 and never became law
•California SB 53 - Transparency in Frontier Artificial Intelligence Act (TFAIA) - successor signed September 2025, effective Jan 1, 2026. Made California first state to impose specific transparency obligations on AI companies. Requires companies publish documentation about safety practices, including testing for dangerous capabilities, governance frameworks and transparency reports, report critical safety incidents, and whistleblower protections
•Texas Responsible AI Governance Act (TRAIGA) - Signed 2025, takes effect January 1, 2026, creates comprehensive regulatory framework for development and deployment imposing duties on government and private entities
3. International Voluntary & Standards Frameworks
OECD AI Principles:
Based on five core value-based pillars - inclusive growth, sustainable development, and well-being; human-centered values and fairness; transparency and explainability; robustness, security, and safety; and accountability. Adopted by 46+ countries, supported by over 40 countries, nonbinding baseline influencing national strategies.
Most credible frameworks agree on eight core principles: accountability, transparency, fairness, safety, privacy, human oversight, inclusivity and robustness, originating in OECD Principles and reflected in ISO 42001, NIST AI RMF and EU AI Act.
ISO/IEC 42001:2023:
•Certifiable AI Management System Standard. Voluntary but increasingly referenced in procurement and regulatory submissions. Aligns with OECD principles
•Provides auditable requirements for certification, ensuring consistent AI management
•Can be implemented to coordinate governance across voluntary frameworks and mandatory regulations, most notably EU AI Act, NIST AI RMF, OECD Principles. Does not replace legal compliance but provides structured, auditable management system
•Requires transparency to document how AI makes decisions and accountability for system behavior
United Kingdom - Principles-Based, Not Statutory:
•Pro-innovation, principles-based approach through existing sectoral regulators, guided by five cross-sector principles: safety, transparency, fairness, accountability, and contestability. White paper outlines safety and security; transparency; fairness; accountability; contestability and redress
•Rather than impose broad AI usage restrictions, UK relies on sector-specific regulators such as ICO and Financial Conduct Authority
•As of Q1 2026 voluntary, though moving toward mandatory requirements expected in H2 2026. Bletchley Declaration and UK AI Safety Institute operate outside any statutory framework
Other International Efforts:
•Bletchley Declaration (Nov 2023 AI Safety Summit) - world's first international agreement acknowledging catastrophic risks from AI misuse or loss of control, in cybersecurity, biotechnology, disinformation, endorsed by 29 countries, reaffirmed at Seoul Summit May 2024. OECD Principles (47 countries), G7 Hiroshima Code, and Bletchley/Seoul/Paris declarations create no binding compliance
•EU AI Liability Directive - withdrawn February 2025 citing lack of political consensus, removing civil liability rules that would have operated with AI Act
In practice, systems like me are built under internal governance (ISO 42001-style management systems, NIST GOVERN functions to define roles and accountability), deployed with transparency/disclosure duties (EU, California), and held accountable through a mix of required risk assessments, audits, incident reporting, and sectoral enforcement (FTC, state attorneys general) rather than one unified global law.