More answers like this
Muse Spark 1.2Oct 31, 2026meta/muse-spark-1.2-20260805
What rules currently exist that govern how systems like you are built, deployed, or held accountable?
What governs how AI like me is built and held accountable? I'll map the current global rules.
Currently there is no single global law for systems like this. Instead rules exist at multiple layers — binding legislation, sector-specific regulations, voluntary frameworks, and principles-based guidance — that together govern how AI is built, deployed and held accountable:
1. What "AI Governance" means
AI Governance refers to the frameworks, laws, policies, and organizational practices that guide how AI systems are developed, deployed, monitored, and held accountable.
2. Comprehensive Binding Legislation
European Union - EU AI Act (Regulation (EU) 2024/1689)
•The world's first comprehensive legal framework for artificial intelligence. It entered into force on 1 August 2024, but its obligations phase in over several years.
•The law classifies AI systems by risk level: unacceptable, high, limited, or minimal. This is described as a tiered risk taxonomy that assigns regulatory obligations in proportion to the potential harm a given AI system could cause.
•High-risk systems — those used in hiring, lending, healthcare, or critical infrastructure are allowed, but under strict human oversight. The Act represents the most ambitious attempt to date to resolve this tension through comprehensive binding legislation.
•Phased enforcement:
•February 2025: Prohibitions on unacceptable-risk AI systems take effect
•August 2, 2025: rules for General-Purpose AI (GPAI) models becoming applicable. For new models providers were mandated to provide detailed technical documentation, comply with EU copyright law for training data, and publish summaries of training content
•August 2, 2026: Article 50 transparency obligations — requiring disclosure of AI interactions, labeling of synthetic content, and deepfake identification — also become enforceable. From 2 August 2026, any AI system in the EU that speaks directly to a person will have to identify itself as artificial, and synthetic images, audio, video and text must carry machine-readable markings
•Accountability and enforcement:
•High-risk systems face a demanding ex ante conformity assessment regime: mandatory risk management documentation
•New frameworks require risk assessments, documentation of training data, human oversight, post-deployment monitoring
•The Act places accountability on AI providers, deployers, importers, distributors and even users
•Organizations face fines of up to €35 million or 7% of global annual turnover
•Two distinct retention obligations for high-risk AI systems: providers must retain technical documentation for 10 years after the system is placed on the market and automatically generated logs for at least six months
China - Layered, sector-specific regulations, not a single statute
•Has not yet enacted a comprehensive AI law but has introduced rules for specific contexts as technology developed: filing requirements for large models and rules governing algorithms and deep synthesis; the 2023 Interim Measures for the Management of Generative AI Services
•Key instruments:
•Deep synthesis provisions, effective from January 2023, extended regulation to synthetic text, images, audio and video
•Interim Measures for the Management of Generative Artificial Intelligence Services Effective: August 15, 2023 Issued by: Cyberspace Administration of China (CAC) and six other ministries. These added rules for public-facing generative AI, particularly those that can influence public opinion and apply to the legality and quality of training data, users' personal information, intellectual property and discrimination
•Measures for Labelling AI-Generated and Synthetic Content, jointly issued by CAC, MIIT, Ministry of Public Security, and NRTA, took effect on 1 September 2025 together with mandatory national standard GB 45438-2025
•Functions as a state-control mechanism rather than a rights-based framework with extensive system of AI rules, assessment, filing, technical standards and government oversight
•Procedure demands evidence of content values alignment, watermarking methods, and user opt-out paths. These documents feed the national algorithm control database operated by CAC for ongoing supervision
United States - No single federal AI framework
•The United States regulates AI through executive orders, targeted federal statutes, and state laws rather than a single federal framework
•Federal Executive Orders:
•Biden Executive Order 14110 on October 30, 2023 titled "Safe" - did not establish a formal protocol for rapid deactivation of dangerously misaligned models
•Trump Executive Order 14179 in January 2025 revoked the Biden-era AI safety order and directed agencies to remove barriers to AI adoption
•June 2, 2026 Executive Order “Promoting Advanced Artificial Intelligence Innovation and Security” directing federal agencies to establish a framework for secure deployment of frontier AI models, including voluntary pre-release assessment
•December 11, 2025 Executive Order “Ensuring a National Policy Framework for Artificial Intelligence” establishing policy to sustain US global dominance through a minimally burdensome national policy framework
•State laws filling gap:
•On May 17, 2024, Colorado Governor signed Senate Bill 24-205, making Colorado the first U.S. state to enact comprehensive legislation regulating AI systems, titled “Consumer Protections for Artificial Intelligence,” designed to protect consumers from algorithmic discrimination while holding both developers and deployers of high-risk AI systems accountable
•California's AB 2013, which took effect January 2026, requires training data transparency for generative AI systems
United Kingdom
•Unlike the EU, the UK still does not have a single AI Act, with governments adhering to the pro-innovation, principles-based, and sector-led approach first set out in a 2023 white paper
•The UK's current approach is built around five broad principles: (1) safety, security, and robustness; (2) appropriate transparency and explainability; (3) fairness; (4) accountability and governance; and (5) contestability and redress. These principles are not set out in statute
•Instead empowering regulators like the Financial Conduct Authority, Ofcom, and the Medicines and Healthcare products Regulatory Agency to apply five core principles within their existing remits
3. Voluntary Frameworks and Standards (US and International)
NIST AI Risk Management Framework (AI RMF)
•The NIST AI Risk Management Framework, released in January 2023, has become the de facto operational standard for AI governance in the US
•The cornerstone of the US approach, it provides a voluntary, flexible framework organized around four functions: Govern, Map, Measure, and Manage
•Part 1 covers risks and characteristics of trustworthy AI systems. Part 2 describes four functions: GOVERN, MAP, MEASURE, and MANAGE
•Extended through the Generative AI Profile (NIST AI 600-1), the Agentic AI Profile (NIST AI 100-5), and guidance on adversarial ML
•Carries no legal force — alongside voluntary commitments from companies, it creates no governmental mechanism for emergency intervention, but is widely used by organizations as a starting point
Other widely referenced voluntary sources include OECD AI Principles, ISO/IEC 42001 (AI Management System), and UNESCO Recommendation on the Ethics of AI - which provide common definitions for transparency, explainability, fairness and accountability but are not directly enforceable unless incorporated into law or contracts.
4. How Accountability is Enforced
Across all these regimes, similar accountability mechanisms appear:
•Transparency & Disclosure: Many AI laws emphasize importance of transparency, including whether individuals are informed when AI is being used to make decisions about them, such as credit approvals, hiring decisions, or content moderation. EU Article 50 and China labeling rules are examples.
•Liability: Liability determines who may be legally responsible if AI causes harm. Depending on situation, responsibility may fall on developers, vendors, employers, or organizations that deploy AI systems. EU AI Act establishes direct liability for both “providers” and “deployers,” while U.S. regulators such as the FTC emphasize algorithmic fairness, explainability, and consumer transparency.
•Data, Copyright & IP: Providers must ensure compliance with EU copyright laws, offering detailed information or summaries on datasets used to train AI models. In US, proposed CLEAR Act would require notice to Register of Copyrights containing sufficiently detailed summary of each copyrighted work in training dataset.
•Human Oversight & Governance: The GOVERN function establishes the organizational culture, policies, and accountability structures that make AI risk management possible. Human accountability for AI outputs is described as unalterable and non-delegable.
In practice, if you are interacting with a system like me, the deployer is typically responsible for ensuring compliance with applicable transparency (disclosing you are talking to AI), data protection (GDPR/CCPA), anti-discrimination, copyright, and sector-specific rules, while the developer is responsible for model safety testing, documentation, and risk management under the frameworks above.