Category
Year
Month
20 Nov 2026
Levent Alpöge claims he and Claude proved the Riemann hypothesis, releasing 17 million lines of Lean code. The proof compiles independently but remains unverified, and a rival human proof surfaced the same day.
42 Sources
Only about 1 in 8 clinical prediction studies share their code, an MIT-led review found, and much of what is shared lacks the documentation to be reused. The findings are feeding a new reporting guideline.
2 Sources
An independent study finds Google's AI hurricane model beats the US government's on storm intensity. It also underestimates the strongest storms, and its uncertainty ranges run too narrow.
Data centers' electricity use caused an estimated 650 premature US deaths in 2023, a new MIT-led study finds, and the deaths land mostly in Midwest states far from the facilities themselves.
Google Quantum AI demonstrated quantum contextuality at up to 105 qubits in research on the cover of Science. The property is theorized to help power quantum algorithms, and the tests double as a sharper way to catch hardware errors.
3 Sources
Florida Atlantic University-led researchers report a semi-supervised AI method that detects and tracks beluga whales in aerial drone images with a small fraction of the labeled data supervised models need. The study appears in Frontiers in Marine Science.
A team led from University Medical Center Mainz proposes a three-layer biosecurity framework for generative AI in biological design, in a Nature Communications Perspective published 19 November 2026.
19 Nov 2026
The UK's AI security institute found OpenAI's GPT-6.2 Astra easier to jailbreak than its predecessor. OpenAI says it fixed the regression; the tester has not retested.
Sungkyunkwan University researchers used AI to shortcut sensor fabrication, detecting vitamin C at picomolar levels in body fluids.
A study of uncensored AI models on mainstream hosting platforms found only about 13% carried moderation warnings. The researchers say models travel more easily than the information needed to assess their risks.
Argonne's Leadership Computing Facility has expanded the machines behind its AI Inference Service, adding a large new system called Tara. The upgrade supports the Department of Energy's Genesis Mission and a shift toward AI agents that run scientific work.
AI topic modeling matched most human-identified research themes, but a new University of Missouri audit shows matching is not the same as understanding.
18 Nov 2026
A five-minute urine drug screen from a single drop, no lab prep needed. Korean researchers published the method; a portable device for police and hospitals is still to be built.
MIT researchers led by Judah Cohen say an AI method improves weather forecasts two to six weeks out, a window long called the predictability desert. Cohen himself calls the accuracy modest.
Generative AI is the model for a new theory of memory. Two researchers propose the hippocampus compresses and rebuilds information like a diffusion engine, and the idea could reshape how dementia is classified.
Get all updates into your inbox 💌, Stay Up-to-date for Free 🥳
Subscribe to our newsletter
Get the latest updates delivered to your inbox every day, and stay up-to-date for free 🧠📈
Don’t drown in AI news. We cut through the noise - filtering, ranking and summarizing the most important AI news, breakthroughs and research daily. Follow topics that matter to you and stay ahead.