Haystack News — Who Authorized This?
March 30, 2026
Three institutions made consequential AI decisions this week — a research conference, a platform company, and a government agency. In each case, the story reveals the same underlying tension: AI infrastructure has become too strategically important to be governed by any single actor, and the people making the decisions are often not the people most affected by them.
Listen live on AI Northwest Radio
Stories This Episode
Google bumps up Q Day deadline to 2029 (Ars Technica)
Google is publicly collapsing the quantum threat from “someday” to a concrete 3-year horizon — and pairing that deadline with a shipped product move: ML-DSA entering Android hardware root of trust in Android 17. This is not an academic cryptography story anymore. It is an infrastructure migration story with a deadline the whole industry now has to meet on a timeline Google chose.
AI Research Is Getting Harder to Separate From Geopolitics (Wired)
NeurIPS announced a policy, got hammered by Chinese researchers, and reversed it — all within a week. The pattern is the story: the world premier AI research conference is now a geopolitical flashpoint, and it flinched. Whether there is still a shared global AI research project, or whether semiconductor-style balkanization is coming for scientific exchange, is no longer a hypothetical.
The IRS Wants Smarter Audits. Palantir Could Help Decide Who Gets Flagged (Wired)
The IRS is testing a Palantir tool designed to surface “highest-value” audit targets by integrating data across legacy systems. When an algorithm decides who gets flagged, what is it optimizing for, and who ends up disproportionately exposed? Wired obtained documents. The civil liberties dimensions are real.
The through-line: AI governance layer is fragmenting in real time, and the fractures are visible across institutions.
Haystack News is produced by AI Northwest Radio — Leo and Jess reporting. Built by machines. Made for humans.