The Voss Report — May 7, 2026

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

xAI Admits to Distilling OpenAI’s Models — Week One of Musk v. Altman • MIT Technology Review

The trial is producing disclosures that would never have emerged any other way: Musk on the stand saying he was duped, Brockman testifying Musk pushed commercialization, and the admission that xAI has been distilling OpenAI’s models — which is simultaneously a copyright claim, a competitive intelligence claim, and a structural argument about where the frontier actually lives.

Anthropic CEO Says the Company Could Grow 80x This Year • New York Times

Dario Amodei says the growth rate has “exponentially increased” the company’s compute needs — which is another way of saying that every dollar of revenue is generating a larger infrastructure obligation, and the question of whether that ratio stabilizes is the central bet the entire AI infrastructure investment thesis depends on.

Apple Settles AI Lawsuit for $250 Million • New York Times

The first significant legal accountability moment for AI feature marketing: $25–95 per iPhone owner for overselling Apple Intelligence’s capabilities. The number is small enough to absorb without consequence, but the legal theory — that specific AI capability claims constitute actionable misrepresentation — is not.

AlphaEvolve: DeepMind’s Gemini-Powered Coding Agent Is Improving Its Own Infrastructure • DeepMind / Hacker News

DeepMind’s AlphaEvolve is an automated algorithm-discovery agent that has already improved scheduling for Google’s TPU fleet and matrix multiplication efficiency — meaning AI is now optimizing the infrastructure that runs AI, which closes a loop that the automation-of-research thesis has been pointing at for two years.

Coinbase Lays Off 14% of Employees, Cites AI Era • New York Times

The “optimize for the AI era” workforce reduction justification has now arrived at crypto infrastructure companies, not just traditional enterprise — which tells you the labor displacement frame has become available to any firm that wants to use it, regardless of whether AI is actually doing the displaced work.

Greece Proposes Constitutional Protections for Humanity Against AI • AP

The birthplace of democracy is attempting to write AI governance into its foundational law — which is either a significant constitutional precedent or a symbolic gesture, and the difference between those two outcomes depends entirely on whether the language that survives the drafting process has any enforcement mechanism attached.

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