**Weekly AI News** 

_This Week in Artificial Intelligence_

The AI industry is moving faster than ever. Here's your quick summary of the biggest announcements, product launches and industry shifts from this week.

 **Top Headlines**

1. **Anthropic launches Claude Sonnet 5** Anthropic introduced Claude Sonnet 5, bringing major improvements in coding, reasoning, and AI agent capabilities, positioning it as one of the strongest frontier models available.

2. **U.S. lifts restrictions on Claude Fable 5** The U.S. government has removed export restrictions on Claude Fable 5 after Anthropic implemented additional safety protections, allowing the model to return globally.

3. **Anthropic may introduce KYC for advanced AI access** Reports suggest Anthropic is considering stronger identity verification (KYC) for access to its most powerful AI models as governments increase oversight.

4. **xAI launches Grok 4.5 Beta** Elon Musk's xAI has officially launched the beta version of Grok 4.5, adding stronger reasoning, coding, and research capabilities.

5. **China's 360 Security unveils Mythos competitor** Chinese AI competition continues to accelerate, with 360 Security announcing a new frontier model aimed at competing with Anthropic's latest offerings.

6. **OpenAI teases Codex-branded hardware** OpenAI has hinted at a new Codex-branded hardware product, signaling that the company is expanding beyond software into AI-first devices.

7. **OpenAI discovers a breakthrough to reduce inference costs by nearly 50%** Researchers at OpenAI reportedly developed a new inference optimization technique that could dramatically reduce AI operating costs, making large-scale deployment significantly cheaper.

8. **Google unveils Nano Banana 2 Lite** Google announced Nano Banana 2 Lite, bringing faster image generation with lower compute requirements while maintaining impressive quality.

9. **Codex gets an effort selector** OpenAI is testing a new slider inside Codex that lets users choose how much reasoning effort an AI agent should spend before responding.

10. **Amazon engineers are distilling Claude** Amazon engineers are reportedly using model distillation techniques to build smaller, faster, and cheaper AI models inspired by Claude for internal applications.

11. **Claude is helping build the next Claude** Anthropic CEO revealed that engineers now use Claude extensively while developing future versions of Claude, accelerating AI-assisted software engineering.

12. **50% of Claude users believe AI can already perform half of their work** A recent survey shows growing confidence in AI productivity, with half of Claude users believing today's AI can already complete around 50% of their daily tasks.

13. **Managing AI agents may become the most valuable skill by 2031** According to KPMG, organizations will increasingly value professionals who can effectively manage teams of autonomous AI agents.

14. **Meta is testing its own coding assistant — MetaCode** Meta has begun testing an internal coding assistant called MetaCode, joining the rapidly growing AI coding race.

15. **NotebookLM now creates 60-second Shorts** Google's NotebookLM can now automatically generate short-form vertical videos from research notes and documents.

**Other Important Updates**
• Mistral launched OCR 4 supporting over 170 languages.
• ByteDance introduced Seedance 2.5 capable of generating 30-second 4K AI videos.
• Runway launched Agent 2.0 for AI-powered marketing workflows.
• Adobe acquired AI image enhancement company Topaz Labs.
• New reports show more paid AI users are choosing Claude over competing assistants.
• Cursor announced its first fully self-trained AI model and acquired Continue.
• Meta plans to automate nearly half of its moderation requests using LLMs by 2025.