International AI Policy: EU AI Policy

Here is an update on European Union AI Policy:

• The EU AI Act entered its main enforcement phase on August 2. Transparency requirements now cover areas such as AI-generated or manipulated content, certain human-facing AI systems and disclosure of AI interaction. European Commission overview

• Regulatory attention is moving from preparation to evidence. Businesses need documentation showing how systems were classified, tested, monitored and governed—not merely general responsible-AI policies. European Commission enforcement update

• Vendors are building compliance into products. Anthropic’s August 14 text-watermarking announcement illustrates how provenance and machine-readable disclosure features are becoming product requirements. Anthropic

• Some high-risk-system deadlines were extended. However, organizations should still complete inventories and governance work because transparency obligations and other provisions are already enforceable.

Regulation and Provenance: Rule to Product Feature

The EU transparency obligations provided the clearest regulatory timeline. On August 6, enforcement and disclosure were the headline. By August 13, feeds emphasized watermarking, machine-readable provenance, and compliance automation. By August 20, provenance had become a product-design and user-experience issue, alongside teen-specific safeguards, state legislation, and concerns about how persistent or connected assistants handle sensitive context. The evolution matters because compliance is moving into the model output, application interface, content pipeline, and audit record.

MANAGEMENT CONSEQUENCE  Do not treat watermarking as a complete provenance system. Retain the source model, prompt or task identity, user or agent identity, processing history, approvals, edits, distribution channel, and final disclosure decision.

Source text: Raw news 8-20.docx — EU AI Policy. AI_Newsletter_V1_Issue_73.docx — How the Major Stories Evolved.