AI capability developments, regulatory frameworks, investment, and the governance contest over artificial intelligence. Published every Friday.
The EU Digital Omnibus package represents a strategic political decision to prioritise industry readiness over regulatory enforcement, creating a 16-24 month window where high-risk AI systems can be d
Read briefing →OpenAI released GPT-5.5 on 2026-04-23, marking a significant advancement in agentic AI capabilities with enhanced coding and multi-tool task execution. This release coincides with OpenAI diversifying
Read briefing →OpenAI released GPT-5.5 on 2026-04-23, marking a significant advancement in agentic AI capabilities with enhanced coding and multi-tool task execution. This release coincides with OpenAI diversifying
Read briefing →Re-issued 2026-04-24 per AD-2026-04-24l-instance-1. Original publish shipped 5 developments; K0-pass2 lab evidence produced 14.
Read briefing →The EU AI Act standards vacuum represents the most immediate governance risk as high-risk compliance deadline approaches without sufficient harmonised standards
Read briefing →EU AI Act Standards Vacuum enables compliance arbitrage until August 2026
Read briefing →Anthropic's double opsec failure in five days — Claude Mythos leaked via CMS and KAIROS/BUDDY/undercover-mode exposed via npm — reveals systemic internal controls failure at the lab with the largest DoD-adjacent AI footprint, exactly as the GSA 'any lawful use' clause threatens to codify no-safety-restrictions as standard federal AI procurement.
Read briefing →China's state formalisation of AI tokens as an economic settlement unit and OpenAI's compute reallocation from video synthesis to robotics pre-training are this week's two structural signals — geopolitical AI architecture is now being built in terminology committees as much as in data centers.
Read briefing →