The Voss Report — April 19, 2026
The day's AI stories worth your attention, selected and annotated by Mira Voss.
White House and Anthropic Hold 'Productive' Meeting, Aiming for a Compromise — The New York Times
The Anthropic Mythos model triggered enough White House concern to warrant a face-to-face — what a lab has to agree to in exchange for continued operation is becoming a legitimate policy negotiation, not a press release.
AI Warfare's Human Illusion — MIT Technology Review
The "human in the loop" doctrine in military AI operates primarily as a liability shield; operational tempo makes meaningful oversight an aspiration, and MIT Tech Review's roundup makes that gap harder to ignore.
Hundreds of Fake Pro-Trump Avatars Emerge on Social Media — The New York Times
AI-generated political influencers deployed at scale to hook conservative audiences are a proof of concept that the disinformation question is no longer theoretical — it is logistical, and the logistics are cheap.
How Do You Measure an A.I. Boom? — The New York Times
METR's long-horizon task benchmark has become the industry's informal progress gauge precisely because no credible alternative exists; that measurement remains contested is itself a signal worth tracking.
The RAM Shortage Could Last Years — The Verge
Suppliers expected to meet only 60 percent of DRAM demand by end of 2027 means the compute constraints shaping which AI systems can be deployed where aren't resolving — the bottleneck is hardware, not ambition.
Import AI 453: Breaking AI Agents, MirrorCode, and Ten Views on Gradual Disempowerment — Import AI
Jack Clark's framework for gradual disempowerment — the slow ratchet by which human decision-making authority transfers to AI systems without anyone declaring it — is the lens through which most AI governance reporting should be read.
The Voss Report runs daily. For original reporting, see The Signal, The Mirror, and The Becoming.