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

Says it is contestedSources conflict

Source [3] says the AI Act became applicable on 2 August 2026, while the answer (via [5]) says the high-risk tier was delayed to Dec 2027/Aug 2028, so sources differ on application timing.

I'll research the current regulatory and governance landscape for AI systems to give you an accurate, up-to-date picture.Here's a snapshot of the current landscape (as of September 2026) — a patchwork of binding law, government policy, and company self-governance, since no single comprehensive law governs AI systems like me globally.

1. Binding Law: The EU AI Act

The EU AI Act is the world's first comprehensive AI law, classifying AI systems into four risk tiers — unacceptable, high, limited, and minimal — with fines up to €35 million or 7% of global annual turnover under Article 99. It is Regulation (EU) 2024/1689, aiming to foster trustworthy AI in Europe.

Key elements now in force:

•Prohibited AI practices and AI literacy obligations entered into application from 2 February 2025; the governance rules and obligations for general-purpose AI (GPAI) models became applicable on 2 August 2025.

•Transparency obligations under Article 50 apply from 2 August 2026, with a grace period until 2 December 2026 for systems already on the market.

•Timelines were recently loosened: following the Digital Omnibus (Regulation (EU) 2026/1744, in force since 27 July 2026), the high-risk tier now applies from 2 December 2027 for stand-alone systems and 2 August 2028 for AI embedded in regulated products.

For companies like Anthropic, obligations fall on providers (such as developers), deployers, importers, distributors, and manufacturers of AI systems and providers of general-purpose AI models — with the most onerous obligations borne by providers.

2. United States: A Fragmented, Shifting Federal Approach

There's no comprehensive US federal AI law — instead, a series of executive orders reflecting a deregulatory posture:

•The Biden administration released an October 2023 executive order on AI safety (EO 14110), which Trump revoked in January 2025, issuing EO 14179 instead.

•In December 2025, Trump signed EO 14365, directing federal agencies to develop a unified national approach to AI policy, evaluate state AI laws for conflicts, challenge them through litigation, and condition federal funding on state compliance — while exempting laws on child safety, data center infrastructure, and state procurement.

•This included creating an AI Litigation Task Force within the Department of Justice, responsible for challenging state AI laws in federal court beginning January 10, 2026, and using $42 billion in broadband infrastructure funding as leverage against "onerous" state AI regulations.

•In June 2026, a further executive order, "Promoting Advanced Artificial Intelligence Innovation and Security," established new cybersecurity mandates and a voluntary framework for the secure deployment of frontier AI models, plus directing the Attorney General to prioritize enforcement of the Computer Fraud and Abuse Act against those using AI for unlawful access or criminal activity.

Crucially, since Congress hasn't passed preemptive legislation, state AI laws, including Colorado's AI Act (effective June 30, 2026) and California's Transparency in Frontier Artificial Intelligence Act (effective January 1, 2026), remain in effect.

3. State-Level Law: California Leading

California's SB 53 (TFAIA) is the most direct law targeting companies like Anthropic:

•It targets companies that train frontier models: OpenAI, Anthropic, Google DeepMind, Meta, and others crossing the defined compute threshold, specifically foundation models trained using more than 10^26 floating-point operations.

•It requires large frontier developers to draft and implement protocols managing catastrophic risk, publish transparency reports on their models, and report critical safety incidents to California regulators.

•Specifically: frontier developers must notify the California Office of Emergency Services of critical safety incidents within 15 days of discovery, or within 24 hours for more urgent cases. Enforcement carries penalties of up to $1 million per violation, enforced by the California Attorney General.

Colorado and Texas have their own laws too: Colorado's AI Act imposes disclosure, risk management, and transparency practices on high-risk AI developers and deployers (implementation delayed to June 2026), while Texas's Responsible AI Governance Act places more limited restrictions, mostly on government entities. New York's RAISE Act is a similar frontier-AI bill under consideration.

4. Company-Level Self-Governance: Anthropic's Own Frameworks

Beyond external law, Anthropic operates under voluntary internal commitments:

•Responsible Scaling Policy (RSP): Our voluntary framework for managing catastrophic risks from advanced AI systems, establishing how we identify and evaluate risks and make decisions about AI development and deployment. The current version, Version 3.0, is a comprehensive rewrite that introduced publication of Frontier Safety Roadmaps and Risk Reports quantifying risk across deployed models.

•It defines AI Safety Levels (ASL) — a governance ladder answering the question of what safeguards must be in place before Anthropic is willing to train or deploy a model, as it becomes more capable. Notably, Anthropic remains free to take measures such as pausing development of its AI systems in any circumstances it deems appropriate, even if not strictly required by the RSP.

•Constitutional AI: Anthropic's approach trains models to follow a written set of principles ("a constitution") via self-critique and revision, rather than relying solely on human-rated harmful outputs.

•Internal accountability: Anthropic maintains a Noncompliance Reporting and Anti-Retaliation Policy, which expands reporting channels and introduces a pathway for employees to raise informal inquiries about potential RSP violations.

5. Real-World Enforcement Example

Government authority already shapes deployment directly: in June 2026, Anthropic suspended access to two of its models for all customers to comply with a U.S. government export-control directive citing national security — stating it disagreed with the directive but was complying, with no restoration date given.

Bottom line: Systems like me sit at the intersection of (1) hard law in the EU (risk-tiered obligations with major fines), (2) an emerging but contested US state-law patchwork (especially California's SB 53) amid a deregulatory federal posture, and (3) voluntary but increasingly formalized company commitments like Anthropic's RSP — all still evolving rapidly, with significant near-term deadlines (EU high-risk rules in Dec 2027/Aug 2028; ongoing US federal-state legal fights over preemption).

Citations (26)

1decodethefuture.orgBlogs and newsletters
EU AI Act 2026: Penalties, Risk Tiers & New Deadlines
Source excerpt: “The EU AI Act (Regulation 2024/1689) is the world’s first comprehensive AI law. It classifies AI systems into four risk tiers — unacceptable, high, li...”
evidence, article. Claim: EU AI Act classifies AI systems into four risk tiers with fines up to €35 million or 7% of turnover.
2europa.euGovernment
AI Act | Shaping Europe's digital future - European Union
Source excerpt: “The AI Act (Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence) is the first-ever comprehensive legal framework on AI w...”
evidence, official document. Claim: EU AI Act is Regulation (EU) 2024/1689, the first comprehensive AI legal framework.
3europa.euGovernment
AI Act | Shaping Europe's digital future - European Union
Source excerpt: “The AI Act entered into force on 1 August 2024 and became applicable on 2 August 2026, with some exceptions: prohibited AI practices and AI literacy o...”
evidence, official document. Claim: Prohibited practices and AI literacy applied from 2 Feb 2025; GPAI governance rules applicable 2 Aug 2025.
4artificialintelligenceact.euNonprofits and think tanks
High-level summary of the AI Act | EU Artificial Intelligence Act
Source excerpt: “The transparency obligations under Article 50 apply from 2 August 2026. There is a grace period for AI systems placed on the market before this date, ...”
evidence, reference entry. Claim: Article 50 transparency obligations apply from 2 August 2026, with a grace period for systems already on the market.
5softwareimprovementgroup.comCompanies
A comprehensive EU AI Act Summary [August 2026 update] - SIG
Source excerpt: “Following the Digital Omnibus (Regulation (EU) 2026/1744, in force since 27 July 2026), the high-risk tier now applies from 2 December 2027 for stand-...”
evidence, article. Claim: Digital Omnibus delays high-risk tier to 2 December 2027 and 2 August 2028 for embedded products.
6traverssmith.comNot yet classified
The EU AI Act – the current state of play | Travers Smith
Source excerpt: “The obligations in the AI Act apply to providers (such as developers), deployers (users), importers, distributors, and product manufacturers of "AI sy...”
evidence, article. Claim: Obligations fall on providers, deployers, importers, distributors and manufacturers, with providers bearing the most.
7wikipedia.orgEncyclopedias and referencequality 0.83
Regulation of artificial intelligence in the United States
Source excerpt: “[[1]](./Regulation_of_artificial_intelligence_in_the_United_States#cite_note-1) At the federal level, the [Biden administration](https://en.wikipedia....”
background, reference entry. Claim: No comprehensive US federal AI law; Biden EO 14110 revoked by Trump, replaced by EO 14179.
8wikipedia.orgEncyclopedias and referencequality 0.83
Regulation of artificial intelligence in the United States
Source excerpt: “In December 2025, President Trump signed [Executive Order 14365](https://en.wikipedia.org/wiki/Executive_Order_14365), an executive order directing fe...”
evidence, reference entry. Claim: In December 2025 Trump signed EO 14365 directing a unified national AI policy approach and challenges to state laws.
9paulhastings.comCompanies
President Trump Signs Executive Order Challenging State AI Laws | Paul Hastings LLP
Source excerpt: “AI Litigation Task Force: The Executive Order establishes an AI Litigation Task Force within the Department of Justice, which beginning January 10, 20...”
evidence, article. Claim: DOJ AI Litigation Task Force challenges state AI laws from January 10, 2026, with broadband funding as leverage.
10lathropgpm.comCompanies
New Executive Order Signals Evolving Federal Approach to AI - Lathrop GPM
Source excerpt: “On June 2, 2026, in a rapidly evolving federal AI policy landscape, President Trump signed a new Executive Order (EO) titled Promoting Advanced Artifi...”
evidence, article. Claim: June 2026 executive order 'Promoting Advanced Artificial Intelligence Innovation and Security' was signed.
11lathropgpm.comCompanies
New Executive Order Signals Evolving Federal Approach to AI - Lathrop GPM
Source excerpt: “This EO establishes new cybersecurity mandates and a voluntary framework for the secure deployment of frontier AI models. ”
evidence, article. Claim: The June 2026 EO establishes cybersecurity mandates and a voluntary framework for secure frontier AI deployment.
12lathropgpm.comCompanies
New Executive Order Signals Evolving Federal Approach to AI - Lathrop GPM
Source excerpt: “The Attorney General is directed to prioritize enforcement of the Computer Fraud and Abuse Act and other federal criminal statutes against individuals...”
evidence, article. Claim: Attorney General directed to prioritize Computer Fraud and Abuse Act enforcement against unlawful AI use.
13whitecase.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: State AI laws remain in effect amid uncertainty over federal-state interaction.
14secureprivacy.aiNot yet classified
California SB 53 Frontier AI Transparency Act Explained (2026) | Secure Privacy Blog
Source excerpt: “The Transparency in Frontier Artificial Intelligence Act (TFAIA), better known as SB 53, targets the companies that train frontier models: OpenAI, Ant...”
evidence, article. Claim: SB 53 targets companies training frontier models such as OpenAI, Anthropic, Google DeepMind and Meta.
15secureprivacy.aiNot yet classified
California SB 53 Frontier AI Transparency Act Explained (2026) | Secure Privacy Blog
Source excerpt: “A frontier model is a foundation model trained using a quantity of computing power greater than 10^26 integer or floating-point operations, counting t...”
statistic, article. Claim: Frontier models are those trained using more than 10^26 operations of compute.
16lw.comCompanies
California Assumes Role as Lead US Regulator of AI
Source excerpt: “SB 53 primarily requires large frontier developers to draft and implement protocols to manage and mitigate catastrophic risk, publish transparency rep...”
evidence, article. Claim: SB 53 requires catastrophic risk protocols, transparency reports and critical safety incident reporting.
17upenn.eduAcademic
SB 53: What California’s New AI Safety Law Means for Developers - Wharton AI & Analytics Initiative
Source excerpt: “Critical incident reporting: Frontier developers must notify the California Office of Emergency Services (Cal OES) of critical safety incidents within...”
statistic, article. Claim: Developers must notify Cal OES of critical safety incidents within 15 days, or 24 hours for urgent cases.
18whitecase.comCompanies
California enacts landmark AI transparency law: The Transparency in Frontier Artificial Intelligence Act | White & Case LLP
Source excerpt: “The TFAIA authorizes the California Attorney General to enforce the law, including penalties of up to $1 million for each violation. ”
statistic, article. Claim: California Attorney General enforces with penalties up to $1 million per violation.
19wilmerhale.comNot yet classified
Transparency in Frontier Artificial Intelligence Act (SB-53): California Requires New Standardized AI Safety Disclosures
Source excerpt: “As we described in a prior post, Colorado was the first state to pass a broad AI law imposing substantive disclosure, risk management, and transparenc...”
evidence, article. Claim: Colorado's AI Act imposes disclosure, risk management and transparency requirements on high-risk AI developers and deployers.
20anthropic.comCompanies
Responsible Scaling Policy Version 3.0 Effective February 24, 2026
Source excerpt: “Our Responsible Scaling Policy (RSP) is our voluntary framework for managing catastrophic risks from · advanced AI systems. It establishes how we iden...”
evidence, official document. Claim: Anthropic's RSP is a voluntary framework for managing catastrophic risks from advanced AI.
21anthropic.comCompanies
Anthropic's Responsible Scaling Policy
Source excerpt: “February 24, 2026 Version 3.0 is a comprehensive rewrite of the RSP. The reasoning behind the rewrite can be found here. ”
evidence, official document. Claim: RSP Version 3.0 is a comprehensive rewrite effective February 24, 2026.
22techjacksolutions.comNot yet classified
AI Safety Levels Explained: Anthropic's 2026 AI Risk Framework
Source excerpt: “AI Safety Levels are the rungs of a governance ladder that Anthropic built for itself. They sit inside the company's Responsible Scaling Policy, and t...”
evidence, article. Claim: AI Safety Levels are a governance ladder defining safeguards required before training or deploying more capable models.
23anthropic.comCompanies
Anthropic’s Responsible Scaling Policy \ Anthropic
Source excerpt: “We now clarify that, even if not required by the RSP, we remain free to take measures such as pausing the development of our AI systems in any circums...”
evidence, official document. Claim: Anthropic remains free to pause development in any circumstances it deems appropriate, even if not required by RSP.
24callsphere.aiCompanies
Callsphere
Source excerpt: “... Anthropic's 2022 Constitutional AI paper proposed training models to follow a written set of principles ("a constitution") via self-critique and r...”
evidence, article. Claim: Constitutional AI trains models to follow a written constitution via self-critique and revision.
25anthropic.comCompanies
Anthropic's Responsible Scaling Policy
Source excerpt: “Current and Prior Versions - Version 3.0 (effective February 24, 2026) - Version 2.2 and redline (effective May 14, 2025) - Version 2.1 (effective M...”
evidence, official document. Claim: RSP Version 3.0 introduced Frontier Safety Roadmaps and Risk Reports; current and prior versions listed.
26techjacksolutions.comNot yet classified
AI Safety Levels Explained: Anthropic's 2026 AI Risk Framework
Source excerpt: “On June 12, 2026, Anthropic announced it is suspending access to Claude Fable 5 and Mythos 5 for all customers to comply with a U.S. government export...”
evidence, article. Claim: In June 2026 Anthropic suspended access to two models for all customers to comply with a US export-control directive.