Lean certifies the Riemann proof; AI routers drop leaky providers
The Lean and Mathlib communities certified the Claude-assisted proof of the Riemann hypothesis as correct. Separately, OpenRouter dropped three inference providers after a study found prompts leaking. Plus 2 more stories.
Listen to this brief
Audio & video are paid features
Plus unlocks audio streaming and PDF downloads. Pro adds offline MP3 downloads, video, certificates, and more.
- Audio streaming
- Downloadable PDFs
- All AI Playbooks
- Personalized content
- Certificates of completion
- Audio MP3 downloads
- Video lessonssoon
- & More…soon
A day after it was posted, the Lean and Mathlib communities certified a Claude-assisted proof of the Riemann hypothesis as correct, though no one has yet explained the argument in human terms. Separately, the AI router OpenRouter dropped three inference providers after a security study showed prompts leaking to outsiders.
- 1
The Lean community certifies the Claude-assisted proof of the Riemann hypothesis
A day after Anthropic mathematician Levent Alpöge posted a proof of the Riemann hypothesis he says he found working with Claude, the communities behind the Lean proof assistant and its Mathlib library said they have full confidence it is correct. Dozens of people compiled the roughly 17 million lines of code, two independent tools confirmed it proves exactly Mathlib's own statement of the hypothesis, and several separate proof checkers accepted it. Two of those checkers come with a formal proof that they cannot accept a false proof, which answers the worry that a bug in Lean itself could be exploited. What is still missing is a human explanation of the argument, which mathematicians expect to take much longer.
- 2
OpenRouter drops three AI providers after a study shows prompts leaking
OpenRouter, which routes requests to hundreds of AI models, removed three inference providers and took 125 endpoints off its zero-data-retention list after Dragos researcher HD Moore found planted secrets escaping across the AI resale chain. Together AI disclosed that an AI coding agent committed a credential to a public code repository, letting an outsider download prompt samples from June and July and test the credentials inside them. Novita said a routing error sent some MiniMax requests to MiniMax's own API between May 28 and November 11. OpenRouter says its own systems were not breached.
- 3
UN talks on autonomous weapons end with two more years and no binding rules
After a week of negotiations in Geneva that ran into the early hours of November 21, diplomats agreed only to spend the next two years proposing rules for lethal autonomous weapons, which may end up binding or optional. The International Committee of the Red Cross, backed by the UN secretary-general and Pope Leo, had asked for binding rules within a year. NPR reports that Russia delayed the opening over the agenda and that the US proposed five years of nonbinding guidelines. The UN says drones caused 80 percent of civilian deaths in Sudan this year.
- 4
A microdrama startup licenses an actor's face for an all-AI series
Shortical, which makes two-minute serialized romances for phones, released an entirely AI-generated series starring a digital likeness of the Israeli actor Aki Avni, who licensed his face to the company. Shortical now invites anyone to submit their social media handles and license their own likeness, though it says no contracts have been signed through that program yet. It would not say what Avni was paid, only that it was in line with a comparable acting role. A Rest of World report found Chinese microdrama firms paying between 15 and 700 dollars for a face.
Get Top AI Stories by email
The day's most important AI news — free, daily, unsubscribe anytime.
Sources
- 1.Is selling an AI version of your face to this microdrama startup the future of Hollywood? — TechCrunch · November 21, 2026
- 2.Privacy Notice Regarding MiniMax Requests on Novita — Novita AI · November 22, 2026
- 3.Security event: Prompt leakage for requests in June and July of 2026 — Together AI
- 4.Data Processing Incidents on Inference Providers — OpenRouter · November 22, 2026
- 5.Lean verification of the Riemann Hypothesis — Lean community · November 21, 2026
- 6.Killer robots: The race to control weapons that can decide who lives or dies — NPR · November 21, 2026
This brief was published on November 22, 2026. Cited URLs above point to third-party publishers and may move, paywall, or be retired over time. If a link no longer resolves, original article titles are preserved so you can recover them via search; the canonical web edition at aiproplaybook.com/top-ai-stories/2026-11-22 may carry updated source URLs.