Skip to content

Updates log

This page aggregates content-level material updates. Code-level changes live in the Git history.

2026-06

Section titled “2026-06”
  • 2026-06-30 — China review completed through the cutoff date: expanded the AI-agent Opinion and TC260-005 analyses; corrected the generative-AI filing model to distinguish national service filing from downstream local registration; added CAC’s 868-service / 530-application implementation totals; and added Shanghai’s government-procurement supervisory-agent trial measure as a sector-specific subnational practice case.
  • 2026-06-30 — United States review completed: added NSPM-11 and Colorado’s psychotherapy-AI restrictions; recorded the Colorado AI Act’s entry into force; and added FTC practice under section 5 and the TAKE IT DOWN Act.
  • 2026-06-30 — European review completed: changed the Digital Omnibus from a pending proposal to an adopted amendment after the Council’s 29 June approval; corrected the AI Act timetable; added the Scientific Panel, Advisory Forum, Germany’s implementation bill and an Italian Garante application case.
  • 2026-06-28 — Site-wide bilingual review through 28 June 2026. English became the default edition; Chinese remains available at matching locale routes. All 158 dated content snapshots now carry the same review date.
  • 2026-06-18 — Added Google DeepMind’s AI Control Roadmap to the company safety-framework analysis, distinguishing system-level agent control from the model-level Frontier Safety Framework.
  • 2026-06-10 — Added the European Commission’s final Code of Practice on Transparency of AI-Generated Content, including its relationship to AI Act article 50 and the 2 August 2026 application date.
  • 2026-06-02 — Added US Executive Order 14409, covering classified cyber-capability benchmarking, voluntary pre-release access for covered frontier models, and the express rejection of mandatory model licensing.

2026-05

Section titled “2026-05”
  • 2026-05-19 — Added TC260-005 Ethics-Safety Guidelines for AI Applications 1.0.
  • 2026-05-08 — Added China’s Implementation Opinion on the Standardised Application and Innovative Development of AI Agents, covering identity, permissions, interoperability and lifecycle safety.
  • 2026-05-07 — Updated the EU AI Omnibus entry to reflect the political agreement and revised high-risk-system dates.

2026-04

Section titled “2026-04”
  • 2026-04-23 — Frontier-lab subpages filled out: Anthropic, OpenAI, Google DeepMind, ByteDance, and DeepSeek now each have substantive analysis across all five standalone subpages (Usage Policy, Model Card, Safety Framework, Transparency Report, Red-team and Evaluation Disclosures), approximately 1,000–1,300 lines per company. A unified top-bar component (SiteHeader) now runs across the homepage and all Starlight internal pages; search and the ⌘K shortcut are wired to pagefind.
  • 2026-04-22 — The corporate-practice axis expanded to 13 companies (added NVIDIA for the infrastructure layer; added a global dimension to the ByteDance–TikTok page; added a retrospective on DeepSeek’s “January 2025 moment”). In-depth analysis of Anthropic RSP v3 (2026-02) and interpretation of Google DeepMind FSF v3 (2026-04).
  • 2026-04-20 — Site launched. Initial topic coverage: risk classification, frontier GPAI, data and training, synthetic-content labelling (four topics × three jurisdictions). Top-level rule coverage: China’s five-tier hierarchy; the US federal layer (including the Trump 2025 series of executive orders); EU AI Act / GDPR / DSA primary-source index. Subnational coverage: Shenzhen; Colorado / California SB 53 / Texas TRAIGA / NYC Local Law 144; Spain AESIA / France CNIL.