The brief · 19 February 2026

The AI brief, 2026-02-19

Another round of health-AI news from hospitals and labs, while Washington tears itself apart over who gets to regulate what. Plus two discoveries that remind us AI is also good at finding things we weren't even looking for.

2 min read J / K to navigate

Weill Cornell launches AI program for medicine

Another medical school moving to structure its AI work instead of leaving it scattered across one-off projects. Research, student training, patient care: all folded under one roof, with ethics and equity billed as guiding principles. Whether this yields real tools or just a lecture series remains to be seen.

practitioners › The program includes a dean's lecture series and an internal grants program, launching February 19, 2026.

Weill Cornell Medicine →

HEALWELL AI and WELL Health Launch WELLTRUST

This platform identifies clinical trial-eligible patients directly within Canadian clinics, with their explicit consent. Nothing changes for you immediately, but it points to where medical AI actually earns its keep: not diagnosing on your behalf, but speeding up research behind the scenes.

practitioners › WELLTRUST runs on DARWEN AI, HEALWELL's patient-identification engine, deployed across WELL Health's clinic network in Canada.

HEALWELL AI / WELL Health →

AI hunts for a rare-earth-free magnet for your EV

Electric motors currently rely on rare-earth magnets, a resource dominated by China and exposed to geopolitical tension. US researchers used AI to screen tens of thousands of magnetic materials and narrow the field to a handful of promising high-temperature candidates.

practitioners › University of New Hampshire: 67,000 magnetic materials screened, 25 compounds shortlisted as potential rare-earth alternatives.

University of New Hampshire →

A Hawaiʻi algorithm learns to find direction in noise

This isn't a product you'll be using tomorrow. But the same method that helps trace a neutrino's path can also clean up a medical scan or stabilize a machine learning model, because the underlying math problem is the same: finding a reliable direction in noisy 2D data.

practitioners › Published February 6 in AIP Advances by a team at the University of Hawaiʻi at Mānoa, using a Frobenius-norm-based method, with applications targeted in imaging, neutrino detection, and ML.

AIP Advances / University of Hawaiʻi at Mānoa →

Super PACs are now waging an AI ad war in the midterms

Two camps are now bankrolling competing campaign ads on AI regulation, including Jobs and Democracy PAC and Leading the Future. The debate has left the technical realm and become a campaign issue in several races, including New York. The federal-vs-state fight may end up decided at the ballot box as much as in Congress.

practitioners › Ad campaigns launched in the New York congressional primary and other races, published February 19, 2026.

Reuters →

US States Split Along Party Lines Over Trump's AI Executive Order

December's executive order directed the DOJ to challenge state AI regulations deemed overly burdensome. Three months in, how states respond tracks which party holds power locally, not the actual substance of the laws being targeted. The practical upshot: what's permitted for professional AI use now depends more than ever on which state you're in.

practitioners › DOJ executive order (December 2025): targeted lawsuits against state AI laws deemed "burdensome," uneven enforcement across states since February 2026.

rollcall.com →