AI news · 10 stories

AI news for Nov 22, 2026

Lean community verifies Claude-assisted Riemann proof · Signal blocks mobile AI agents from capturing message screens · UN talks extend autonomous-weapons discussions by two years

News card: Lean community verifies Claude-assisted Riemann proof
Lean community

Lean community verifies Claude-assisted Riemann proof

The Lean and Mathlib communities said November 21 that they verified Levent Alpöge’s Riemann Hypothesis proof, which he says he developed with Claude. Independent checks of the formal statement and proof strengthen the evidence for AI-assisted mathematical discovery.

Read at Lean community
News card: Signal blocks mobile AI agents from capturing message screens
Signal

Signal blocks mobile AI agents from capturing message screens

Signal announced November 20 screenshot protections against AI agents on iOS 27 and supported Android versions. The measures limit screen capture of private messages, while Signal warns they cannot prevent every route to application data.

Read at Signal
News card: UN talks extend autonomous-weapons discussions by two years
Associated Press

UN talks extend autonomous-weapons discussions by two years

UN-backed talks in Geneva concluded November 21 with an agreement to continue autonomous-weapons discussions for two years. The compromise did not secure a firm commitment to binding rules, leaving restrictions on AI-powered weapons unresolved.

Read at Associated Press
News card: OpenAI details defenses against adversarial distillation
OpenAI

OpenAI details defenses against adversarial distillation

OpenAI described stronger reasoning-extraction defenses and coordinated campaign detection on November 20. In its internal adaptive evaluation, extraction of earlier reasoning fell from 54% for original GPT-6 Astra safeguards to 0.2% for updated GPT-6.2 Astra safeguards, addressing one route for copying capabilities without protections.

Read at OpenAI
A glowing antenna broadcasting neon signal rings
Google

Google brings Duplex business calls into Tasks in Gemini

Google explained November 20 how its Gemini-based Duplex stack now powers business calls through Tasks in Gemini. AI disclosure and merchant opt-outs accompany broader appointment and request handling, extending assistants into real-world phone coordination.

Read at Google
News card: MIT introduces physics constraints for stable chaotic-system models
MIT News

MIT introduces physics constraints for stable chaotic-system models

MIT reported November 20 a machine-learning approach that enforces energy dissipation to keep chaotic-system predictions bounded. The research aims to avoid physically impossible runaway outputs in applications such as climate and turbulent-flow modeling.

Read at MIT News
News card: AWS launches cross-region HyperPod Global Inference Gateway
Amazon Web Services

AWS launches cross-region HyperPod Global Inference Gateway

AWS introduced SageMaker HyperPod Global Inference Gateway on November 20, routing requests across registered clusters and regions through one managed endpoint. Automatic failover and capacity-aware routing help teams use spare GPUs without maintaining separate client-side routing systems.

Read at Amazon Web Services
A glowing antenna broadcasting neon signal rings
Amazon Web Services

AWS Quick adds coordinated agents and persistent web workflows

AWS updated Quick on November 20 with conversational flow building, multi-agent orchestration and work that continues after the browser closes. Deterministic steps and reusable skills help teams turn individual routines into repeatable business processes.

Read at Amazon Web Services
News card: Apple releases Depth Pro 2 for consistent image and video depth
Apple

Apple releases Depth Pro 2 for consistent image and video depth

Apple released Depth Pro 2 code and weights alongside research submitted November 19, using one network to estimate metric depth from images and video at native resolution. The approach targets both sharp object boundaries and consistency between frames, improving the building blocks for spatial vision.

Read at Apple
News card: NVIDIA details Dynamo improvements for reinforcement-learning rollouts
NVIDIA

NVIDIA details Dynamo improvements for reinforcement-learning rollouts

NVIDIA described November 19 how Dynamo balances cached prompts and worker load for reinforcement-learning rollouts. ModelExpress transfers weight changes and supports recovery, helping reduce repeated computation and training interruptions.

Read at NVIDIA

Get the next episode first

Shorts and deep dives on AI system design land on @AI.JoinDev first. Subscribe, or follow the RSS feed.