📰 Key AI news & updates

Key AI news & updates

Key AI news & updates

1. OpenAI released the open‑weight safety reasoning models “gpt‑oss‑safeguard‑120B / 20B” for public use.
   What it means: Developers can now use open‑weight models to enforce custom safety / content / moderation policies via chain‑of‑thought reasoning, raising the bar for transparency and customizability in safety classification.

2. OpenAI launched its new AI‑powered web browser “ChatGPT Atlas” on macOS (with Windows, iOS, Android to follow).
   What it means: AI is being embedded more deeply into the core browsing experience, enabling browsing + summarization + agentic interactions in‑context rather than switching apps, shifting how users will interact with web content.

3. India proposed amendments to its digital / intermediary / media rules requiring visible labeling + metadata traceability for generative / synthetic media.
   What it means: Regulatory scrutiny is tightening in India: generative AI content will face mandated transparency (disclaimers / marking synthetic content), signaling stronger oversight of deepfakes or AI‑generated media.

4. The NITI Aayog (via India’s government) released a roadmap titled AI for Inclusive Societal Development, focusing on applying AI for informal workforce / inclusive growth.
   What it means: India is formalizing strategic alignment of AI with societal inclusion, workforce reskilling, and bringing AI to underserved workers or informal sectors; sets priorities for public, industry & academic collaboration.

5. Meta Platforms (Meta) reported Q3 earnings and announced significantly increased AI infrastructure / compute spending for 2026, despite a large tax charge.
   What it means: Big tech remains doubling down on AI, increasing compute investment even through macro pressures — suggests AI infrastructure arms race continues, which could feed into downstream tools / platforms.

One tool to try

Try using ChatGPT Atlas browser — for integrated browsing + AI assistant + agent mode, so you can browse, ask questions about pages, fill forms or perform tasks directly from the the browser context.

One repo / chart

Check out the updated version of the International AI Safety Report 2025: First Key Update: Capabilities and Risk Implications (arXiv preprint) for recent updates on AI capabilities, risks and policy implications