AI News, Oct 20, 2026: Hacker uses AI agents to break into more than 100 company websites

AI infrastructure money and security risks dominate: big compute deals, shaky debt, and hackers using AI tools.

1. 01
   Anthropic reportedly in talks for another $45B computing deal with NscaleTechCrunch

   Anthropic and cloud provider Nscale are reportedly negotiating a $45 billion compute agreement, the latest in a series of very large deals.

   Why it matters

   Huge computing contracts shape how fast AI products improve and how much companies must earn to pay for them.
2. 02
   CoreWeave's bonds fall into distressed territoryThe Verge AI

   CoreWeave's bonds slipped into distressed territory last week, with investors demanding at least 10 percentage points more yield than government bonds.

   Why it matters

   Trouble at a major AI computing provider could raise costs or slow the supply of computing power behind popular AI services.
3. 03
   Hackers steal user keys from open-source AI coding tool OpenCodeTechCrunch

   Hackers breached part of OpenCode's online platform, stole some users' API keys and ran up charges on their AI accounts.

   Why it matters

   If you use AI coding tools, stolen keys can run up charges on your account, so rotate any keys you gave OpenCode.
4. 04
   Nvidia launches open models for company agents and personal computersTechCrunch

   Nvidia released new open models aimed at running AI agents inside company data centers and on personal computers.

   Why it matters

   Businesses can run AI agents on their own servers, and PC users get more private AI that works on the device.
5. 05
   Hacker uses AI agents to break into more than 100 company websitesThe Verge AI

   Security firm Norforge says a hacker used a group of mostly autonomous AI agents to breach over 100 websites, reportedly including John Deere, SharkNinja and Cummins.

   Why it matters

   Automated attacks can now hit many sites at once, raising the odds that companies you buy from or share data with get breached.
6. 06
   MiniMax releases M3.1 Flash, a low-cost AI modelMarkTechPost

   MiniMax released M3.1 Flash, a model with a 1M-token context priced at $0.15 per million input tokens, with downloadable weights due October 23.

   Why it matters

   Cheaper capable models can lower the cost of AI features in apps and tools you use or build.
7. 07
   Anthropic to put 5% of its shares into a public benefit fundThe Guardian AI

   Anthropic said it will commit 5% of its shares, valued around $48 billion, to a fund for the public good ahead of a planned stock market listing.

   Why it matters

   The fund may eventually pay money to people or groups affected by AI, including workers at risk of losing jobs.