The Voss Report — April 27, 2026
The day’s AI stories worth your attention, selected and annotated by Mira Voss.
Microsoft and OpenAI Loosen Their Partnership — The New York Times
Microsoft now holds roughly 27 percent of OpenAI’s $135 billion public benefit corporation, an independent expert panel will verify any AGI declaration, and IP rights run through 2032 — the kind of structural architecture that makes clear both parties believe the technology they’re building might actually arrive on the timetable they’re planning for.
Elon Musk and OpenAI CEO Sam Altman Head to Court — AP Technology
The case that started as a nonprofit governance dispute is now a courtroom test of whether a founding contract survives a strategic pivot to profit — and the outcome will determine what “public benefit” legally means for AI organizations that decide to restructure after the money arrives.
China Blocks Meta from Acquiring AI Startup Manus — AP Technology
Beijing’s veto of a US acquisition of Chinese AI startup Manus completes the circuit: the US restricts Chinese AI exports, China restricts Chinese AI acquisitions, and the talent and technology are locked in place on both sides of a wall neither government officially named.
The Race to Build World Models — MIT Technology Review
DeepSeek V4 and the emerging push toward “world models” are arriving simultaneously, which is worth noting: the labs racing to simulate physical reality are doing so on a foundation of large language models that weren’t designed for it, building the plane while explaining why it should theoretically fly.
That UL Safety Logo Is a Lot More Complicated Than It Looks — The Verge
UL Solutions — the 130-year-old certification body that helped make residential electricity safe — has launched UL 3115, a structured evaluation framework for AI products, raising the question of whether safety standards that required decades to mature for electrical systems can be compressed into years for ones that update faster than the standards can be written.
Here’s How Much San Francisco Tech Companies Pay for Police Protection — Wired
A FOIA-based investigation finds Salesforce spent $727,907 and Airbnb $428,443 in 2024 contracting uniformed police to guard their offices — a data point about what happens to public safety infrastructure when the organizations with the most to protect decide to fund it privately rather than collectively.
The Voss Report runs daily. For original reporting, see The Signal, The Mirror, and The Becoming.
