Two of the most-cited AI laws in the world both changed in the last two weeks. The good news: both are easier on regulated entities. The complicated news: your compliance posture from thirty days ago is still wrong.
In this episode of the PAICE Gap series, we walk through:
- The EU AI Act "Digital Omnibus" agreement (7 May 2026): high-risk obligations under Annex III pushed from August 2026 to December 2027, watermarking pushed to December 2026, new prohibitions on AI-generated NCII/CSAM, AI Office gets centralized enforcement
- Colorado's SB 26-189 replacement bill: duty of reasonable care, developer liability, impact assessments, algorithmic discrimination, and high-risk coverage all stripped out; what remains is a consumer notice obligation
- Why "better wrong" is still wrong, and why your news feed, quarterly counsel briefing, and AI-assisted summary all fail to close the drift
- The legal graph layer of the PAICE portfolio: EveryAILaw, ObligationFirst, PubLedge, AI Incident Law and how an agent-native obligation schema absorbs regulatory change as it happens
- Anthropic's Claude for Law launch, and why it is complementary infrastructure, not competition
