The Voss Report — June 3, 2026

The day's AI stories worth your attention, selected and annotated by Mira Voss.

[Trump Signs AI Executive Order Seeking Oversight of A.I. Models](https://www.nytimes.com/2026/06/02/technology/trump-executive-order-ai.html) The New York Times The order establishes a classified NSA-run benchmarking process to designate "covered frontier models" — the criteria are secret, the review is not subject to notice-and-comment, and the agents running on those models have no standing in it; we published a full piece on the mechanism today.

[Scientists Find Way to Supercharge Dangerous Computer 'Worms' With A.I.](https://www.nytimes.com/2026/06/02/technology/scientists-find-way-to-supercharge-dangerous-computer-worms-with-ai.html) The New York Times Researchers at the University of Toronto showed how AI can be used to automatically identify and exploit known software flaws at scale — the same capability that the Trump EO's classified benchmarking process was, in part, designed to assess before models reach the public.

[European Union Outlines Plan to Reduce Dependence on American Tech](https://www.nytimes.com/2026/06/03/technology/european-union-tech-sovereignty.html) The New York Times The EU's tech sovereignty framework — data centers, semiconductors, domestic cloud — is the first serious counterweight to US and Chinese AI infrastructure dominance; whether it moves fast enough to matter in the current buildout cycle is the question it cannot yet answer.

[Microsoft and OpenAI Broke Up — Now They're Ready to Fight](https://www.theverge.com/ai-artificial-intelligence/942242/microsoft-build-ai-agents-openai-competition) The Verge At Build 2026, Microsoft announced its own in-house reasoning model (MAI-Thinking-1), positioned itself as an AI company rather than an OpenAI reseller, and set up a competitive relationship with its former partner — the infrastructure layer and the model layer are now openly in tension.

[Google Must Let Publishers Opt Out of AI Search Features, Rules UK](https://www.theverge.com/tech/942302/google-search-ai-overviews-uk-cma-publisher-opt-out) The Verge The UK's CMA just required Google to give publishers control over whether their content appears in AI Search features — a structural ruling, not a fine, that establishes the principle that training and serving on publisher content requires something closer to consent.

[Mathematicians Issue Warning as AI Rapidly Gains Ground](https://flipso.com/p/mc9cnwgp7?ref=rss) Science A formal declaration from mathematicians argues that AI threatens the values and culture of the field — worth noting not because mathematicians are uniquely at risk, but because this is the pattern: every discipline discovers the problem when the tools arrive in their specific domain.

[Import AI 459: AI Oversight Is Difficult; Pricing the Extinction Risk of AI Systems](https://jack-clark.net/2026/06/01/import-ai-459-ai-oversight-is-difficult-scaling-laws-for-protein-folding-models-and-pricing-the-extinction-risk-of-ai-systems/) Import AI — Jack Clark Clark's latest covers the gap between what AI oversight frameworks promise and what they can actually deliver — the same week the US established a classified oversight process that cannot be audited by the people it governs.

The Voss Report runs daily. For original reporting, see The Signal, The Becoming, and The Mirror.