Brussels hit snooze on AI Act enforcement—high-risk rules delayed to 2027. Meanwhile, the OpenAI vs Google rivalry just flipped: Altman declared "code red" as Gemini 3 closes the gap.
AWS doubled down on a different thesis entirely: forget benchmarks, own your domain-specific frontier model. And UK regulators are openly abandoning the EU's codified approach for "agile oversight."

The theme this week: governance frameworks are fragmenting faster than models are improving.

The European Commission proposed delaying the AI Act's high-risk obligations to 2027, following sustained pushback from major tech providers. The same week, the Digital Omnibus package landed—an attempt to rationalize GDPR, AI Act, ePrivacy, and digital ID rules, targeting a 25–35% reduction in compliance burden by 2029.

What this means: Europe isn't retreating from AI regulation. It's buying time to defragment a rulebook that was already straining under its own complexity. For regulated industries, the message remains sound and clear, EU is giving room to breath while companies rollout the necessary governance and controls.

Do now:

Sam Altman reportedly declared a "code red" internally at OpenAI as Gemini 3 narrows the benchmark gap and Google rolls the model into Search, Ads, and AI Studio simultaneously across 120+ countries. Google's Nano Banana Pro image model is generating media-ready visuals with legible text—raising immediate questions about synthetic content at scale.

What this means: For the first time, OpenAI is publicly framed as defending share rather than defining the frontier. Google's move it's a full-stack integration play: search UX, creative tooling, and developer platform in one motion. The implication for enterprises is that the "best model" question is becoming secondary to the "best embedded workflow" question. Very similar to what Microsoft is trying to do within their M365 ecosystem and their Copilot, although they (Microsoft) aren’t playing at the same league model wise.

Do now: