EU’s cyber agency and research centre are testing Chinese AI models
THE SO WHAT
EU security and science arms quietly benchmarking open-weight Chinese models means model provenance is now a formal risk variable, not just a performance question. If you deploy open models in Europe, assume regulators will soon have an opinion on which families are acceptable for which data classes.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIRevolut to roll out agentic shopping as AI commerce takes off
Banking apps turning into agentic shopping front-ends means checkout, credit, and loyalty all compress into one surface. If you own a consumer brand, assume your next customer touchpoint may be an AI agent inside someone else’s financial app, not your site or app.
Applied AISmall businesses that adopt AI are more optimistic about growth, revenue, and employment
SMBs reporting higher optimism on growth and hiring after adopting AI suggests the near-term story is augmentation, not headcount cuts. If you sell into SMBs, position AI as revenue and capacity expansion rather than cost takeout.
Applied AIAnalysis of 857 releases from nine Chinese AI labs between 2021 and September 2026: just 3.6% disclosed safety results, and only 1.1% did so at launch
If only 3.6% of 857 Chinese AI lab releases disclosed safety results—and just 1.1% at launch—then transparency is still the exception, not the rule. Multinational teams relying on these models need their own eval and governance layers rather than assuming upstream disclosure.
Applied AIInstinct was the buzziest AI agent around — can it survive Muse?
The fact that an invite-only, text-first agent like Instinct could become “buzzy” before being overshadowed by more visual agents shows how fluid this category is. If you’re building agents, assume UX and distribution will be copied quickly—durability has to come from data, integrations, or niche depth.