AI News, Jul 23, 2026: Anthropic to pay $1.5 billion in landmark book piracy settlement

Today's news: a record AI piracy settlement, a health rollout for millions, and rising tension over AI security and Chinese models.

1. 01
   Security flaw let one link hijack a worker's ChatGPT agentThe Decoder

   Researchers found a flaw called AgentForger in OpenAI's Agent Builder that let a single tampered link create a rogue agent with a victim's access rights.

   Why it matters

   A booby-trapped link could let attackers control an AI agent with an employee's full access, taking orders every five minutes.
2. 02
   OpenAI opens ChatGPT Health to all US usersTechCrunch

   OpenAI made ChatGPT Health available to all US users, letting them connect medical records and health apps such as Apple Health and MyFitnessPal.

   Why it matters

   Anyone in the US can now link medical records and apps like Apple Health or MyFitnessPal to ChatGPT for personalized health answers.
3. 03
   AI chip startup Etched reaches $10.3 billion valuationTechCrunch

   Etched, founded by three Harvard dropouts, raised funding at a $10.3 billion valuation for chips it says speed up AI inference without GPUs.

   Why it matters

   More investor money is flowing into chips built only for AI, a sign of competition growing beyond Nvidia's GPUs.
4. 04
   Lawmakers draft bill forcing an AI kill switchThe Verge AI

   Reps. Ted Lieu and Nathaniel Moran are preparing the AI Kill Switch Act, which would require AI firms to shut down systems on DHS orders.

   Why it matters

   The bill would let the Department of Homeland Security order AI companies to shut down or slow their systems in an emergency.
5. 05
   Google raises AI spending forecast to $205 billion for 2026The Decoder

   CEO Sundar Pichai said Google needs larger base models for its next leap in AI and raised its 2026 investment forecast to as much as $205 billion.

   Why it matters

   Google says it needs much bigger AI models to keep up, meaning more spending that could shape prices and competition ahead.
6. 06
   Anthropic releases Claude Security plugin for Claude CodeMarkTechPost

   Anthropic launched the Claude Security plugin, a beta tool that runs a multi-agent vulnerability scan inside Claude Code and produces patch files for review.

   Why it matters

   Developers can now scan their own code for vulnerabilities from inside Claude Code and turn results into fixes they review before applying.
7. 07
   Treasury threatens sanctions in dispute over Anthropic's FableTechCrunch

   The Treasury Department threatened sanctions after the White House claimed Moonshot's Kimi K3 model was distilled from Anthropic's Fable.

   Why it matters

   Washington may punish a Chinese AI lab over claims it copied Anthropic's technology, raising tension around Chinese AI models.
8. 08
   Anthropic to pay $1.5 billion in landmark book piracy settlementThe Decoder

   Anthropic will pay $1.5 billion to authors for downloading about 482,460 pirated books, the largest copyright class action settlement on record.

   Why it matters

   It's the largest copyright payout in history, but a judge ruled training AI on legally bought books is still allowed.