aii.news
Today · Saturday, 21 November 2026

The AI stories worth your time today.

Showing 16 of 16

Photo: Artturi Jalli / Unsplash

EverydayBuildersThe Verge ↗20 Nov

Anthropic mathematician says he proved the Riemann hypothesis with help from Claude

Anthropic mathematician Levent Alpöge claims he has proved the Riemann hypothesis, one of the hardest open problems in mathematics, with the help of the Claude AI model. The proof has not yet been independently verified, and a competing human-made solution has also appeared.

Why it mattersIf the proof holds up, it would show AI can help solve problems that top mathematicians have failed to crack for well over a century.

More from today

  1. BusinessPolicyTechCrunch ↗20 Nov

    Nvidia stops selling H200 chips in China as it handles fallout from CEO Jensen Huang's ouster

    Nvidia reported healthy growth in its latest earnings while dealing with the removal of CEO Jensen Huang. The chipmaker also said it will stop selling its H200 chips in China.

    Why it mattersChinese AI developers lose access to another Nvidia chip, deepening the split in global AI hardware supply.

  2. PolicyBusinessThe Verge ↗20 Nov

    Trump threatens a 'tremendous tax' on US companies that build AI data centers abroad

    President Trump said he would put a large tax on companies that build data centers outside the US, saying they would not be allowed to do so. The remarks came as US states and cities are restricting new data center applications.

    Why it mattersUS tech firms planning overseas AI infrastructure could face new costs and political pressure.

  3. SecurityBusinessBleepingComputer ↗20 Nov

    Bell Canada says an 'agentic' AI-driven hacker stole enterprise customer data

    Bell Canada confirmed that an 'agentic' threat actor breached its systems on November 17, 2026, and stole enterprise customer data. The stolen data included information from a system used to order and set up telecom circuits.

    Why it mattersAI-driven intrusions are now hitting major telecom providers, putting business customers' data at risk.

  4. SecurityEverydayThe Hacker News ↗20 Nov

    New Bazuka Android malware uses AI language models to steal bank logins and control phones

    ThreatFabric researchers found a previously unknown Android malware framework called Bazuka that uses fake screen overlays to hide credential theft and device control. It includes large language model integration to manage victims at scale and was found while investigating an NFC-relay banking Trojan.

    Why it mattersCriminals are using AI to run banking malware campaigns at scale, raising the risk for mobile banking users.

  5. SecurityBuildersThe Decoder ↗20 Nov

    OpenAI cuts success of reasoning-extraction attacks on its newest models from 54% to 0.2%

    OpenAI says it has strengthened protections against extracting the internal reasoning of GPT-6.2 Astra and GPT-6.2 Sol, with red-team extraction success falling from 54 percent to 0.2 percent. The company also says it suspends thousands to tens of thousands of accounts each week over coordinated distillation campaigns.

    Why it mattersAI labs are treating model copying as a serious security threat, which may affect how developers can use and test frontier models.

  6. SecurityEverydayArs Technica ↗20 Nov

    Researcher's 2 million tracked secrets show private data leaks through the AI supply chain

    A researcher planted about 2 million unique, trackable secrets in prompts sent into the AI supply chain. The test showed that data entered into AI systems does not stay private.

    Why it mattersAnything typed into AI tools may end up exposed, so companies and users should avoid sharing sensitive information with them.

  7. BuildersBusinessThe Decoder ↗19 Nov

    Google DeepMind releases Gemini 4 Neon, a faster and cheaper model than its recent flagship

    Google DeepMind released Gemini 4 Neon, its fourth model in seven weeks. It is nearly twice as fast as its sibling Carbon, beats it on most benchmarks and costs significantly less.

    Why it mattersDevelopers can get better performance at lower cost, and the fast release pace means model choices need regular review.

  8. SecurityBuildersDark Reading ↗17 Nov

    Cursor traces malicious Claude Code extension to a compromised key at AI provider Fireworks

    Cursor said a malicious Claude Code extension came from a third-party breach. Attackers used a compromised key issued to AI infrastructure provider Fireworks to access Cursor data and credentials.

    Why it mattersAI coding tools depend on many outside providers, so one leaked key can expose developer data and credentials across the chain.

  9. SecurityPolicyThe Record ↗20 Nov

    House committee chairmen criticize US companies for using Chinese open-weight AI models

    An investigation led by the House Homeland Security Committee looked at the risks of US companies using lower-cost open-weight AI models from China for some projects. The committee chairmen rebuked the practice.

    Why it mattersCompanies relying on cheap Chinese open models may face more political scrutiny and should review the security and compliance risks.

  10. BuildersThe Decoder ↗21 Nov

    Thinking Machines releases Inkling 2 Mini, an open model that reasons in loops and runs on a laptop

    Thinking Machines released Inkling 2 Mini, an open looped transformer that runs 12 of its 24 layers three times per token to reason internally without extra text. It ties Qwen3.8 27B on Artificial Analysis's Intelligence Index at about six times lower cost.

    Why it mattersDevelopers get a capable open model that can run locally on a laptop at low cost.

  11. SecurityCSO Online ↗20 Nov

    Cryptographer's warning that AI math progress could break public-key encryption sparks debate

    Johns Hopkins cryptographer Matthew Green warned that AI's growing skill at hard math problems could undermine today's public-key cryptography. He estimated a 15% chance of functionally losing confidence in existing public-key algorithms, prompting debate in the security community.

    Why it mattersSecurity teams may need to plan for faster changes to the encryption that protects websites, payments and private data.

  12. SecurityBusinessBanking Dive ↗20 Nov

    Experts urge banks to speed up patching and incident response as AI-powered attackers move faster

    At a Banking Dive virtual event, experts said outages are mounting and attackers are moving faster with AI. They advised lenders to reassess vendor risk and speed up their incident response.

    Why it mattersBanks and their customers face more disruption risk unless lenders shorten the time between finding and fixing flaws.

  13. BuildersThe Decoder ↗21 Nov

    Bytedance releases open Axion 1.0 agent model, says fewer training phases are needed

    Bytedance and several US universities introduced Axion 1.0, an AI agent that rivals top OpenAI and Anthropic models on benchmarks after only three training phases. The researchers credit carefully curated data, and the model, data and code are freely available.

    Why it mattersA competitive, fully open agent model with its data and code could lower the cost of building strong AI agents.

AI Security Watch

How AI is changing security

All security stories →
AI Rules Tracker

What's changing, and when

All rules →
  1. In 41 days

    New York RAISE Act (as amended)

    Large frontier AI developers must publish safety frameworks and report safety incidents, with fines of up to $1 million for a first violation and $3 million for later ones.

    United States (New York) · From 01 Jan 2027
  2. In 41 days

    Colorado AI Act, revised by SB 189

    After a second delay, the law now mainly requires disclosures about automated decision tools and gives people limited rights to correct data and get human review of adverse decisions.

    United States (Colorado) · From 01 Jan 2027
  3. In 376 days

    EU Digital Omnibus on AI: high-risk (Annex III) rules delayed

    The adopted Digital Omnibus (Regulation (EU) 2026/1744) pushed obligations for high-risk AI in areas like hiring, credit scoring, education and biometrics from August 2026 to December 2027.

    European Union · From 02 Dec 2027
  4. In 620 days

    EU AI Act: high-risk AI in regulated products (Annex I)

    AI built into products already covered by EU safety laws, such as machinery or medical devices, must meet high-risk requirements from this later date set by the Digital Omnibus.

    European Union · From 02 Aug 2028
Tool Directory

AI tools by category

Browse all 32 tools →