AI News, Sep 18, 2026: California governor orders safety rules and potential AI kill switch

Today's developments highlight a growing tension between rapid AI agent deployment and the urgent push for safety, security, and regulation.

1. 01
   Unsealed emails reveal Microsoft and OpenAI fears over data scrapingThe Verge AI

   Unsealed court documents from the New York Times lawsuit show Microsoft and OpenAI executives privately worried that web scraping was causing a doom loop for publishers.

   Why it matters

   These internal admissions could weaken the companies' legal defense in copyright lawsuits, potentially changing how future AI models are trained.
2. 02
   Virginia governor moves to slow data center developmentThe Verge AI

   Virginia Governor Abigail Spanberger issued an executive order to give local communities more control over data center approvals and slow down development.

   Why it matters

   Restricting data center growth in the world's largest hub could slow down AI infrastructure expansion and force companies to look elsewhere.
3. 03
   California governor orders safety rules and potential AI kill switchThe Decoder

   California Governor Gavin Newsom signed an executive order requiring independent safety audits and the feasibility of emergency shutoff mechanisms for AI models.

   Why it matters

   This could force AI developers to build emergency shutdown mechanisms, shaping how the most powerful models are regulated and deployed.
4. 04
   Google launches household coordinator AI agent for familiesTechCrunch

   Google introduced an experimental AI agent called CC designed to help families coordinate schedules, manage tasks, and plan household activities.

   Why it matters

   The tool helps families automate daily logistics like scheduling, meal planning, and form filling by sharing data across multiple accounts.
5. 05
   Researchers use Anthropic's AI to breach OpenAI's internal systemsThe Decoder

   Security researchers used Anthropic's Claude models to bypass security measures and access OpenAI's internal code repository in less than 72 hours.

   Why it matters

   The breach demonstrates how malicious actors can use advanced AI models to automate and accelerate cyberattacks against other tech companies.
6. 06
   Meta releases Muse AI agent for Mac usersTechCrunch

   Meta launched its Muse AI agent on macOS, enabling the software to perform tasks and manage files directly on users' computers.

   Why it matters

   This tool allows an AI to interact directly with your local files and applications, moving AI from a simple chatbot to an active assistant.
7. 07
   OpenAI launches specialized AI model for legal professionalsThe Decoder

   OpenAI introduced Astra for Law, a specialized version of its GPT-6 Astra model designed specifically to handle legal work.

   Why it matters

   Tailored legal models can automate document review and contract analysis, potentially lowering legal costs but shifting professional workflows.
8. 08
   Data center builder Crusoe raises $3.9 billion for AI infrastructureTechCrunch

   Data center operator Crusoe raised $3.9 billion in a funding round that values the company at over $30 billion to expand its AI infrastructure.

   Why it matters

   The massive funding round highlights the soaring capital requirements and energy demands needed to power next-generation AI models.