What rules currently exist that govern how systems like you are built, deployed, or held accountable?

Says it is contested

I'll research the current regulatory landscape for AI systems to give you an accurate, up-to-date answer.The landscape of AI governance is fragmented, rapidly evolving, and operates on multiple levels simultaneously — international, national, sub-national (state), and voluntary/self-imposed. Here's a breakdown of what currently applies to systems like me:

1. The EU AI Act (the most comprehensive binding law)

Adopted on 21 May 2024, the EU Artificial Intelligence Act lays down harmonized rules on artificial intelligence and is the first-ever comprehensive legal framework on AI worldwide. It uses a graduated, risk-oriented structure that differentiates AI systems by the potential harm they may pose, imposing escalating obligations where risks increase.

Key elements now in force or phasing in:

•Prohibited practices: Certain AI practices have been prohibited under Art. 5 AI Act since February 2, 2025, including social scoring, subliminal manipulation, and real-time biometric remote identification in public spaces. As of December 2026, these prohibitions extend to "nudifier" apps — AI systems that generate or manipulate sexually explicit or intimate images, video, or audio without explicit consent, or that create CSAM.

•Transparency duties: The transparency duties (Article 50) and the AI Office's enforcement powers over general-purpose AI (GPAI) providers took effect exactly as scheduled on 2 August 2026 and were not delayed. This means providers/deployers must disclose AI interaction, mark synthetic or manipulated content in a machine-readable way, inform people about emotion-recognition or biometric-categorization systems, and label deepfakes or certain AI-generated public-interest text.

•GPAI model obligations: Every provider of a general-purpose AI model (like the large language model underlying me) placed on the EU market must maintain detailed technical documentation covering model architecture, training procedures, and performance characteristics; provide documentation to downstream providers; publish a sufficiently detailed summary of training data content using the AI Office's template; and comply with EU copyright law, including respecting text-and-data-mining opt-out mechanisms.

•High-risk system rules: These have been pushed back — high-risk obligations for stand-alone systems (Annex III, covering hiring, credit scoring, education, and critical infrastructure) are deferred to 2 December 2027, and high-risk AI embedded in already-regulated products (Annex I, covering medical devices, machinery, and toys) is deferred to 2 August 2028.

•Penalties: For the deepfake/nudifier prohibitions alone, violations may trigger fines of up to €35 million or 7% of annual worldwide turnover, whichever is higher.

2. U.S. Federal Policy — deliberately light-touch, with a national-security carve-out

There's no comprehensive U.S. federal AI statute. Instead, the current posture is:

•State preemption push: In December 2025, President Trump signed an executive order titled "Ensuring a National Policy Framework for Artificial Intelligence," which establishes a federal policy aimed at addressing the growing number of state-level AI regulations... "to sustain and enhance the United States' global AI dominance through a minimally burdensome national policy framework for AI." It directs the Attorney General to establish an AI litigation task force to challenge state AI laws deemed inconsistent with the Executive Order's language, including on grounds of unconstitutional regulation of interstate commerce and federal preemption. It also uses federal funding as leverage to limit state AI regulation by authorizing federal agencies to condition discretionary grants on states refraining from enacting AI laws deemed inconsistent with the policy. Notably, though, the final text expressly prohibits federal preemption of otherwise lawful state AI laws, including those relating to child safety, AI compute and data center infrastructure, state government procurement and use of AI.

•National security framework: In June 2026, the White House issued an executive order titled "Promoting Advanced Artificial Intelligence Innovation and Security," establishing a voluntary framework permitting developers of certain advanced AI models to voluntarily provide the federal government with model access prior to public release for cybersecurity and national security assessments. It also instructs specified agencies to create an AI cybersecurity clearinghouse to centralize information on AI-related vulnerabilities and threats. Critically, as of now the order does not define "covered frontier model," leaving scope uncertain.

•Because Congress hasn't passed preemptive legislation, existing state AI laws, including Colorado's AI Act and California's Transparency in Frontier Artificial Intelligence Act, will likely not be impacted in the short term.

3. U.S. State Laws — the most concrete accountability mechanisms right now

•California's TFAIA (SB 53), effective January 2026, is the most directly relevant to a system like me. It applies to a foundation model that was trained using a quantity of computing power greater than 10^26 integer or floating-point operations — and requires "large frontier developers" (revenue over $500M) to write, implement, comply with, and clearly and conspicuously publish a frontier AI framework, reviewed and updated at least once per year, describing how they incorporate national and international safety standards and define capability thresholds for catastrophic risk. It also mandates notifying the California Governor's Office of Emergency Services of any "critical safety incidents" within 15 days of discovery and establishes incident reporting mechanisms, creates research initiatives, and provides robust whistleblower protections.

•Colorado AI Act, effective mid-2026, imposes various obligations related to documentation, disclosures, and governance of "high-risk" AI systems—systems that make "consequential decisions."

•Texas RAIGA places narrower categorical limits on AI development/deployment.

4. International Voluntary Commitments

Governments have also pursued soft-law approaches: the G7 Hiroshima Process International Code of Conduct and the Seoul AI Safety Summit's Frontier AI Safety Commitments are voluntary international frameworks that companies like Anthropic have signed onto.

5. Company Self-Governance (like Anthropic's Responsible Scaling Policy)

Since binding law hasn't fully caught up to frontier AI, companies operate under their own voluntary frameworks, which increasingly inform the law rather than the reverse. Anthropic's Responsible Scaling Policy (RSP) is a good example directly relevant to how I'm built and deployed:

•It's explicitly Anthropic's voluntary framework for managing catastrophic risks from its own AI systems. As one analysis notes, it's not a regulation and not an industry standard — it's one company's voluntary framework, published openly, and it has become one of the most-cited reference points in frontier AI governance.

•It defines AI Safety Levels (ASL): ASL-3 is the first tier where strict deployment and security safeguards become mandatory under the Responsible Scaling Policy.

•Version 3.0 (Feb 2026) added Risk Reports — detailed information on the safety profile of models at time of publication, going beyond capabilities to address threat models and active risk mitigations, providing an assessment of overall risk level — plus external review procedures.

•Notably, this voluntary self-governance is starting to get codified: Anthropic's ASL-3 safeguards influenced legislation like California's SB 53, New York's RAISE Act, and the EU AI Act's Codes of Practice, which now require frontier AI developers to publish risk frameworks. Anthropic now maintains a separate compliance document (the FCF) that serves as our compliance framework for SB 53 and other regulatory requirements, distinct from the RSP itself.

•Anthropic also collaborates with government organizations like the UK AI Security Institute and US Center for AI Standards and Innovation (CAISI) for independent testing.

The bottom line

Right now, systems like me sit at the intersection of: binding EU law (transparency, documentation, copyright compliance, and — soon — high-risk rules), a patchwork of binding U.S. state laws (especially California's and Colorado's), a U.S. federal approach that favors deregulation and preemption while carving out national-security review, international voluntary pledges, and company-specific safety policies that often set the de facto standard before law formalizes it. This is a genuinely unsettled area — several of the frameworks above (the EU's Omnibus amendments, the U.S. federal preemption fight, state law enforcement) are still being litigated or negotiated, so specifics could shift again within months.

Citations (29)

1legalnodes.comCompanies
EU AI Act 2026 Updates: Compliance Requirements and Business Risks
Source excerpt: “Adopted on 21 May 2024, the EU Artificial Intelligence Act lays down harmonized rules on artificial intelligence and is the first-ever comprehensive l...”
evidence, article. Claim: The EU AI Act was adopted 21 May 2024 as the first comprehensive legal framework on AI worldwide.
2legalnodes.comCompanies
EU AI Act 2026 Updates: Compliance Requirements and Business Risks
Source excerpt: “The EU’s regulatory model for artificial intelligence is founded on a graduated, risk-oriented structure. Rather than targeting specific technologies,...”
background, article. Claim: The EU AI Act uses a graduated, risk-oriented structure with escalating obligations as risks increase.
3lw.comNot yet classified
AI Act Update: EU Resolves to Change Rules and Extend Deadlines
Source excerpt: “Certain AI practices have been prohibited under Art. 5 AI Act since February 2, 2025, including social scoring, subliminal manipulation, and real-time...”
evidence, article. Claim: Certain AI practices, including social scoring and subliminal manipulation, have been prohibited under Art. 5 since February 2, 2025.
4lw.comNot yet classified
AI Act Update: EU Resolves to Change Rules and Extend Deadlines
Source excerpt: “Effective December 2, 2026, the Agreement extends these prohibitions to “nudifier” applications — AI systems that generate or manipulate sexually expl...”
evidence, article. Claim: From December 2026, prohibitions extend to nudifier apps generating sexually explicit images without consent or creating CSAM.
5softwareimprovementgroup.comCompanies
A comprehensive EU AI Act Summary [August 2026 update] - SIG
Source excerpt: “Meanwhile, the transparency duties (Article 50) and the AI Office’s enforcement powers over general-purpose AI (GPAI) providers took effect exactly as...”
evidence, article. Claim: Article 50 transparency duties and AI Office GPAI enforcement powers took effect on 2 August 2026 as scheduled.
6decodethefuture.orgBlogs and newsletters
EU AI Act 2026: Penalties, Risk Tiers & New Deadlines
Source excerpt: “Depending on the use case, providers or deployers must disclose AI interaction, mark synthetic or manipulated content in a machine-readable way, infor...”
evidence, article. Claim: Providers and deployers must disclose AI interaction, mark synthetic content, and label deepfakes.
7decodethefuture.orgBlogs and newsletters
EU AI Act 2026: Penalties, Risk Tiers & New Deadlines
Source excerpt: “Every provider of a GPAI model placed on the EU market must: maintain detailed technical documentation covering model architecture, training procedure...”
evidence, article. Claim: GPAI providers must keep technical documentation, publish training data summaries, and comply with EU copyright law.
8softwareimprovementgroup.comCompanies
A comprehensive EU AI Act Summary [August 2026 update] - SIG
Source excerpt: “The Omnibus confirms two fixed dates for the Act’s toughest tier: high-risk obligations for stand-alone systems (Annex III, covering hiring, credit sc...”
evidence, article. Claim: High-risk obligations deferred to 2 December 2027 for Annex III and 2 August 2028 for Annex I systems.
9lw.comNot yet classified
AI Act Update: EU Resolves to Change Rules and Extend Deadlines
Source excerpt: “Violations may trigger fines of up to €35 million or 7% of annual worldwide turnover, whichever is higher. ”
statistic, article. Claim: Violations may trigger fines up to €35 million or 7% of worldwide turnover.
10whitecase.comCompanies
State AI laws under federal scrutiny: Key takeaways from the executive order establishing federal AI policy framework | White & Case LLP
Source excerpt: “On December 11, 2025, President Trump signed an executive order titled “Ensuring a National Policy Framework for Artificial Intelligence,” (“Executive...”
evidence, article. Claim: Trump signed a December 2025 executive order on a national policy framework for AI.
11kslaw.comCompanies
New State AI Laws are Effective on January 1, 2026, But a New Executive Order Signals Disruption | King & Spalding
Source excerpt: “The Executive Order directs the Attorney General to establish an AI litigation task force (the “Task Force”) to challenge state AI laws deemed inconsi...”
evidence, article. Claim: The order directs the Attorney General to create an AI litigation task force to challenge state AI laws.
12paulhastings.comCompanies
President Trump Signs Executive Order Challenging State AI Laws | Paul Hastings LLP
Source excerpt: “The Executive Order uses federal funding as leverage to limit state AI regulation by authorizing federal agencies to condition discretionary grants on...”
evidence, article. Claim: The order uses federal funding as leverage by conditioning discretionary grants on states not enacting conflicting AI laws.
13paulhastings.comCompanies
President Trump Signs Executive Order Challenging State AI Laws | Paul Hastings LLP
Source excerpt: “The final text is also narrowed as it expressly prohibits the federal preemption of otherwise lawful state AI laws, including those relating to child ...”
evidence, article. Claim: Final text expressly prohibits preemption of lawful state AI laws on child safety, infrastructure, and state procurement.
14hklaw.comCompanies
Executive Order on Artificial Intelligence Expands Cybersecurity, Federal Oversight | Insights | Holland & Knight
Source excerpt: “The White House on June 2, 2026, issued an executive order (EO), titled "Promoting Advanced Artificial Intelligence Innovation and Security," establis...”
evidence, article. Claim: June 2026 executive order establishes a voluntary framework for pre-release government access to advanced AI models.
15congress.govGovernment
Controlling Advanced Artificial Intelligence: Executive Order 14409 Explained | Congress.gov | Library of Congress
Source excerpt: “The order instructs specified agencies to create an AI cybersecurity clearinghouse to centralize information on AI-related vulnerabilities and threats...”
evidence, official document. Claim: The order instructs agencies to create an AI cybersecurity clearinghouse for vulnerabilities and threats.
16congress.govGovernment
Controlling Advanced Artificial Intelligence: Executive Order 14409 Explained | Congress.gov | Library of Congress
Source excerpt: “The order does not define covered frontier model. ”
evidence, official document. Claim: The order does not define covered frontier model, leaving scope uncertain.
17whitecase.comCompanies
State AI laws under federal scrutiny: Key takeaways from the executive order establishing federal AI policy framework | White & Case LLP
Source excerpt: “While the interaction between federal initiatives and existing state requirements remains uncertain and may vary depending on legislative, regulatory ...”
evidence, article. Claim: Interaction between federal initiatives and existing state laws remains uncertain, so state laws likely unaffected short term.
18whitecase.comCompanies
California enacts landmark AI transparency law: The Transparency in Frontier Artificial Intelligence Act | White & Case LLP
Source excerpt: “... Frontier model: a foundation model that was trained using a quantity of computing power greater than 10^26 integer or floating-point operations. ...”
statistic, article. Claim: California's TFAIA defines frontier models as trained with more than 10^26 operations.
19mofo.comNot yet classified
At the Frontier – California Enacts AI Safety and Transparency Regulation TFAIA (SB 53) | Morrison Foerster
Source excerpt: “TFAIA requires a large frontier developer to write, implement, comply with, and clearly and conspicuously publish on its internet website a frontier A...”
evidence, article. Claim: Large frontier developers must write, implement, and publish a frontier AI framework.
20swlaw.comNot yet classified
California Enacts Landmark AI Safety and Transparency Law - Law Offices of Snell & Wilmer
Source excerpt: “... TFAIA requires frontier developers to notify the California Governor’s Office of Emergency Services (OES) of any “critical safety incidents” relat...”
evidence, article. Claim: Developers must notify California OES of critical safety incidents within 15 days.
21aoshearman.comNot yet classified
California adopts landmark AI law: new transparency, safety, and reporting requirements for frontier model developers
Source excerpt: “The law introduces new obligations for AI developers, establishes incident reporting mechanisms, creates research initiatives, and provides robust whi...”
evidence, article. Claim: The law establishes incident reporting, research initiatives, and whistleblower protections.
22crowell.comNot yet classified
California’s Landmark AI Law Demands Transparency From Leading AI Developers | Crowell & Moring LLP
Source excerpt: “For example, the Colorado Artificial Intelligence Act, which will go into effect on June 30, 2026, imposes various obligations related to documentatio...”
evidence, article. Claim: Colorado AI Act, effective June 30, 2026, imposes documentation, disclosure and governance obligations on high-risk AI.
23techjacksolutions.comNot yet classified
AI Safety Levels Explained: Anthropic's 2026 AI Risk Framework
Source excerpt: “The Responsible Scaling Policy, or RSP, is Anthropic's voluntary framework for managing catastrophic risks from its own AI systems. ”
attribution, article. Claim: Anthropic's Responsible Scaling Policy is a voluntary framework for managing catastrophic risks.
24techjacksolutions.comNot yet classified
AI Safety Levels Explained: Anthropic's 2026 AI Risk Framework
Source excerpt: “This is not a regulation and not an industry standard. It is one company's voluntary framework, published openly, and it has become one of the most-ci...”
attribution, article. Claim: RSP is not a regulation or industry standard but one company's voluntary framework widely cited in governance.
25techjacksolutions.comNot yet classified
AI Safety Levels Explained: Anthropic's 2026 AI Risk Framework
Source excerpt: “ASL-3 is the first tier where strict deployment and security safeguards become mandatory under the Responsible Scaling Policy, which is why it is the ...”
evidence, article. Claim: ASL-3 is the first tier where strict deployment and security safeguards become mandatory under the RSP.
26anthropic.comCompanies
Responsible Scaling Policy Version 3.0 Effective February 24, 2026
Source excerpt: “Risk Reports will provide detailed information on the safety · profile of our models at the time of publication. They will go beyond describing model ...”
evidence, official document. Claim: RSP version 3.0 added Risk Reports detailing model safety profiles.
27mlq.aiCompanies
Anthropic Releases Revised Responsible Scaling Policy 3.0 with Adjusted Safety Commitments | MLQ News
Source excerpt: “[2][3] This success influenced legislation like California's SB 53, New York's RAISE Act, and the EU AI Act's Codes of Practice, which now require fro...”
evidence, article. Claim: Anthropic's safeguards influenced SB 53, New York's RAISE Act, and EU Codes of Practice requiring risk frameworks.
28anthropic.comCompanies
Anthropic’s Transparency Hub \ Anthropic
Source excerpt: “The FCF serves as our compliance framework for SB 53 and other regulatory requirements. ”
evidence, official document. Claim: Anthropic's FCF serves as its compliance framework for SB 53 and other regulations.
29anthropic.comCompanies
Anthropic’s Transparency Hub \ Anthropic
Source excerpt: “We collaborate with government organizations like the UK AI Security Institute and US Center for AI Standards and Innovation (CAISI) for independent t...”
evidence, official document. Claim: Anthropic collaborates with UK AI Security Institute and US CAISI for independent testing.