🤖 White House eyes pre-release AI vetting, Microsoft Agent365 reins in shadow AI
Audio in Mandarin Chinese · English transcript below
⚡ White House wants to pre-approve every model, Microsoft collars shadow AI, Musk slams OpenAI greed, Amex lets AI swipe your card, China lags eight months, image AI funding rockets—and yes, it can still teach you bioweapons.
Today's Top 3 Headlines
- Technology
🤖 White House Eyes Pre-Release AI Model Review
Bloomberg: White House drafting rule for pre-release safety reports on ChatGPT-class LLMs; failures blocked. Compliance costs jump, startups may earmark 20% extra funding for red-team audits & algorithm filings.
Source ↗ - Technology
🤖 AI chatbot gives scientists step-by-step bio-weapon recipe
Nature study: top LLMs detail plague & botulinum weapon recipes in 30 s. Devs must harden refusals & red-team now, or face forced shutdown.
Source ↗ - Enterprise Security
🛡️ Microsoft Agent 365 tames enterprise “shadow AI” in 1 click
Microsoft unveils Agent 365 at Build: central registry & policy guardrails for employee-built gen-AI agents. Brings “shadow AI” under one roof, cutting data-leak & compliance risk while speeding sanctioned tools.
Source ↗
+7 more headlines
- 🤖 Musk’s lawyer grills Brockman: why chase $30B?
- 🤖 AmEx builds Agent Commerce with AI agents + intent contracts; trades stay in-house
- 🤖 Gemini personalization hits HK; AI Plus/Pro/Ultra sync Gmail & YouTube
- U.S. Claims China’s Top LLM Trails by Eight Months
- 🤖 Pinecone Nexus engine beta: 2.8M tokens → 4K
- 🤖 Image LLM gains 34M fans in 28d, downloads surge 6.5×
- 🤖 Karpathy: LLMs still can’t grasp “wash the car” common sense