A regulatory map, not a marketing checklist.

The EU AI Act's obligations on providers of general-purpose AI models begin to apply in full on August 2, 2026. Article 53 sets out documentation and transparency duties. Article 54 covers appointed representatives for non-EU providers. Article 55 adds evaluation, systemic-risk mitigation, and incident reporting for models designated as posing systemic risk.

VerifAI's Compliance Kit is written against the enacted text, not a summary of it. The evidence surface — training documentation, deployment records, incident trail — is structured to answer the questions the AI Office is expected to actually ask.

US federal AI regulation remains sectoral. For banks, SR 11-7 is the operative model-risk framework and now the de facto template for governing generative and agentic systems.

State law is moving faster than federal. Colorado's AI Act (SB24-205, effective 2026) sets consumer-notice and documentation duties for high-risk AI. California and New York are moving on shutdown authority and frontier-model safety reporting. Any US deployment worth defending needs an inventory of models, a documented kill path, and a clear owner.

The UK's AI Regulation White Paper approach places responsibility with existing regulators — the FCA, PRA, MHRA, ICO, and Ofcom — rather than a single AI act. In practice that means an AI system in UK financial services already sits inside SS1/23 model risk expectations, and one in health already sits inside MHRA software-as-a-medical-device.

Japan has stayed lighter-touch than the EU but has converged on the same evidence expectations through the Hiroshima Process and the METI/MIC AI Guidelines for Business. A single evidence pipeline can satisfy EU Article 53 documentation and Japanese transparency expectations at the same time.

China's Interim Measures for the Management of Generative AI Services (in force since 2023, extended in 2025) require algorithm filing, content labelling for AI-generated media, and adherence to socialist core values. Multinationals operating in China maintain a separate compliance track; VerifAI's evidence layer is compatible with the filing structure but does not substitute for local filing itself.

The Australian government has issued a Voluntary AI Safety Standard and is consulting on mandatory guardrails for high-risk AI. VerifAI's compliance surface is designed to satisfy the voluntary standard today and to convert cleanly into mandatory documentation when the guardrails become law.