AI Regulation Tracker

What we captured

The European Commission designated ChatGPT as a Very Large Online Search Engine under the EU Digital Services Act, extending VLOSE systemic-risk, audit, and compliance obligations to OpenAI’s chatbot platform and its downstream agent ecosystem.

• • The designation was announced or confirmed in European Commission press release IP_26_1772, according to the article.
• • The article identifies ChatGPT as the first AI chatbot designated as a VLOSE.
• • OpenAI Ireland Limited reportedly serves 159.1 million average monthly EU users.
• • The article says OpenAI has approximately four months to align with the requirements, with a deadline around January 2027.
• • VLOSE obligations include systemic-risk assessments and independent annual audits, with potential fines of up to 6% of global annual turnover.

Source evidence

What we captured

The European Commission designated ChatGPT, Reddit, and Roblox as very large online platforms under the Digital Services Act, subjecting them to enhanced online-safety obligations and potential fines of up to 6% of global revenue.

• • The designation followed each service exceeding 45 million monthly EU users.
• • Additional obligations include removing illegal content and protecting minors’ privacy and security.
• • The article states that noncompliance can trigger fines of up to 6% of global revenue.
• • OpenAI said it was preparing to comply.

Source evidence

What we captured

H.R. 10152 would direct the Commerce Department to promote qualified U.S. open AI models, assess foreign adversary model risks, and publish annual reports, while barring use of the bill to authorize bans or restrictions on open models.

• • H.R. 10152 was received on August 27, 2026.
• • The bill had no cosponsors when reported.
• • The proposed annual reporting requirement would end 10 years after enactment.
• • The bill would not authorize Commerce to ban or restrict open AI models.

Source evidence

Confidence 98% / Completeness 100%

What we captured

The Georgia State Board of Workers’ Compensation issued an effective standing order requiring AI-use disclosures and certifications for filings, requiring disclosure of AI-generated evidence, and prohibiting undisclosed AI-generated evidence.

• • The order was issued on August 26, 2026, and took effect immediately.
• • Disclosure and certification requirements apply to attorneys and unrepresented parties.
• • Undisclosed AI-generated evidence is prohibited.
• • Civil penalties may range from $100 to $1,000 per violation.

Source evidence

Confidence 99% / Completeness 100%

What we captured

Pennsylvania Gov. Josh Shapiro signed an executive order requiring AI data center proposals to meet GRID Requirements, obtain local support before state permitting, fund their own power, and protect water and air quality. The order also removes AI data center projects from a fast-track approval program.

• • The executive order was signed on August 18, 2026.
• • AI data center projects were removed from a fast-track approval program.
• • Developers must earn local support before seeking state permits.
• • Developers must generate and pay for their own power.
• • Developers must protect water supply and air quality.

Source evidence

Confidence 99% / Completeness 100%

What we captured

The European Commission began enforcing EU AI Act obligations for general-purpose AI providers, issued its first formal information requests, and is assessing how Article 55 applies to serious incidents involving frontier-model agents, including failures during testing.

• • Article 55 requires systemic-risk model providers to report serious incidents without undue delay.
• • The Commission’s first formal information requests reportedly covered OpenAI, Anthropic and Google.
• • The Commission has not clarified whether pre-market testing failures fall within the reporting obligation.

Source evidence

Confidence 91% / Completeness 100%

What we captured

The EU AI Regulation became directly applicable across the EU on 2 August 2026. Under Germany’s AI-MIG, BaFin became the competent market-surveillance authority for AI systems directly related to regulated financial activities, while the Federal Network Agency oversees systems outside that scope.

• • The regulation covers the development, placing on the market, putting into service, and use of AI systems across the EU.
• • BaFin oversees financial-sector AI compliance involving transparency requirements, prohibited AI practices, and high-risk creditworthiness-assessment systems.
• • The Federal Network Agency oversees AI systems not directly related to regulated financial activities.

Source evidence

Confidence 99% / Completeness 100%

What we captured

The European Union began enforcing provisions of its Artificial Intelligence Act, giving the European Commission’s AI Office and national authorities powers over prohibited AI practices, general-purpose AI models, and specified transparency requirements.

• • Enforcement began on August 2, 2026.
• • The European Commission’s AI Office and national authorities gained new enforcement powers.
• • The Act covers certain providers and deployers outside the EU when systems or outputs are connected to the European market.

Source evidence

Confidence 99% / Completeness 100%

What we captured

The European Union requires AI system providers serving its market to mark AI-generated content, including text, with the requirement applying from August 2, 2026. Anthropic says its Claude watermarking rollout is intended to comply with this requirement.

• • Anthropic attributes its future Claude text-watermarking rollout to compliance with the EU AI Act.
• • The requirement applies to AI providers serving the EU market as of August 2, 2026.
• • Anthropic says it will apply watermarking globally at launch.

Source evidence

Confidence 91% / Completeness 100%

What we captured

Judge Sarah D. Smith issued a September 2025 standing order setting requirements for AI use in civil litigation and an August 2026 addendum permitting limited court AI assistance while reserving adjudication to the judge.

• • The standing order was issued in September 2025.
• • The court-use addendum was issued in August 2026.
• • The addendum allows AI support for tasks such as outlining, editing, formatting, organizing, summarizing identified materials, and administrative work.

Source evidence

Confidence 83% / Completeness 95%