TL;DR
China, the United States and European Union are activating three different forms of AI pre-release oversight within 19 days. The rules share an interest in government scrutiny before deployment, but differ on legal force, policy goals and covered systems.
China, the United States and the European Union are activating different forms of government scrutiny for advanced AI systems within a 19-day period, beginning with China’s anthropomorphic-interaction rules on July 15 and continuing with US and EU milestones on August 1 and 2. The clustered deadlines matter because developers serving all three markets face separate approval, evaluation and conformity systems rather than a shared global standard.
China’s Interim Measures for AI Anthropomorphic Interaction Services took effect on July 15, according to the Thorsten Meyer AI dispatch. Issued in April by five agencies, including the Cyberspace Administration of China, the measures extend the country’s existing regulatory approach to human-like AI services such as companion systems and agents. The dispatch says covered providers may face security reviews before public deployment, algorithm registration and continuing government oversight.
In the United States, the dispatch identifies August 1 as the point when measures under Executive Order 14409 take hold. It describes a voluntary framework offering participating frontier-model developers a 30-day government evaluation window based partly on classified criteria. Unlike China’s system, the US mechanism is presented as an incentive-based access process, with trusted-partner status and procurement opportunities used to encourage participation.
The EU AI Act is scheduled to become fully applicable on August 2 under its staged timetable. Its system centers on risk classification, conformity assessments, technical records and post-market monitoring. General-purpose AI models above the law’s systemic-risk threshold face added evaluation and incident-reporting duties. The three systems consequently examine different risks through different legal mechanisms.
Three Gates Close in Nineteen Days
The Pre-Release Regime Goes Global
Same-day-verified · one instinct, three architectures — and none of them binds the open frontier
Anthropomorphic-interaction measures take effect: five agencies extend the CAC approval regime to companion AI and agents.
EO 14409’s classified benchmark and voluntary 30-day pre-release framework harden. NSA designates covered frontier models.
The AI Act becomes fully applicable — the staged rollout that began February 2025 reaches its final station.
Same instinct, three theories of a gate
STEELMAN: THE GATE-SKEPTIC CASE
Pre-release regimes structurally favor incumbents who can afford the process — and none of the three binds an open-weight release from a lab outside its jurisdiction. The gates go up exactly as the fastest-moving part of the frontier walks around them.
The signal: a model can clear all three gates having been evaluated for three almost non-overlapping things — content control, fundamental rights, national security. Jurisdiction is now an architectural property. If your deployment calendar doesn’t carry July 15, August 1, and August 2, it’s a calendar for a market you’re not in.
AI compliance and regulation books
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Global Developers Face Three Tests
The alignment of these dates shows a shared policy instinct: certain AI systems should face government scrutiny before broad public use. It does not amount to regulatory alignment. China focuses on content control and social stability, the EU on fundamental rights and product safety, and the United States on national-security risks through voluntary cooperation.
For developers, jurisdiction is becoming part of system design. A company may need to build market-specific reporting, documentation and incident processes into its release plans. Clearing one jurisdiction’s gate does not establish compliance elsewhere because the tests cover largely different questions. Smaller developers may also face higher relative costs because they have fewer legal and compliance resources than established technology companies.
Three Regulatory Models Converge
China has required security assessments for some public generative AI services since 2023. The dispatch describes a five-step CAC registration process under which regulators can seek design changes before issuing a filing number. It also reports ongoing duties that include a 24-hour incident-reporting period, responses to government information requests within 48 hours and implementation of ordered algorithm changes.
The EU timetable began with prohibited AI practices in February 2025, followed by general-purpose AI obligations in August 2025. The United States has pursued a lighter approach centered on access, evaluation and procurement incentives. The United Kingdom remains outside this set of formal gates, relying instead on a principles-based system administered by sector regulators.
“Same instinct, three theories of a gate.”
— Thorsten Meyer AI dispatch
Deadlines and Coverage Remain Fluid
Several points remain unsettled. The dispatch says a proposed EU Digital Omnibus would delay certain high-risk AI deadlines after European Parliament approval on June 16 by a vote of 423–57, with 174 abstentions. It also says the measure has not completed Council adoption or Official Journal publication, leaving August 2 as the operative date for now.
The practical reach of the US system is also uncertain because participation is voluntary and the central evaluation criteria are classified. It is not yet clear how many developers will participate, which models the National Security Agency will designate, or how strongly procurement incentives will influence release decisions. The source also argues that none of the systems fully binds open-weight models released outside the relevant jurisdiction, but the extent of that gap will depend on enforcement and market-access rules.
August Deadlines Test Enforcement
Attention now moves to the August 1 US milestone and the EU’s scheduled full application one day later. Developers will be watching for NSA model designations, participation details and evidence of how US trusted-partner status affects procurement. In Europe, the immediate questions are whether the Digital Omnibus completes the legislative process and which deadlines, if any, are formally changed.
China’s first weeks of enforcement may also show how authorities interpret anthropomorphic interaction, which companion services and agents fall within scope, and whether regulators request product or algorithm changes. Those decisions will indicate whether the three gates remain separate national systems or begin producing shared compliance practices across global AI companies.
Key Questions
What happened on July 15?
China’s anthropomorphic-interaction measures took effect, extending its AI oversight approach to human-like services, including companion AI and agents, according to the dispatch.
Are the three systems legally equivalent?
No. China uses a state approval and oversight model, the EU requires risk-based conformity processes, and the US framework described by the source is voluntary and incentive-based.
Has the EU’s August 2 deadline already moved?
Not based on the source material. The dispatch says the Digital Omnibus is not yet in force, so August 2 remains operative unless the remaining legislative steps are completed and published.
Do these measures cover every frontier AI model?
No. Coverage differs by jurisdiction, model type and deployment method. The source identifies a possible gap involving open-weight releases from foreign developers, while the US framework’s voluntary status may leave some companies outside government evaluation.
Why do the clustered dates matter?
Companies operating across the three markets may need to satisfy multiple release processes almost simultaneously. The deadlines also show that pre-release AI scrutiny is spreading without producing one global regulatory standard.
Source: Thorsten Meyer AI
