🤖 Hassabis pitches U.S. safety body for frontier AI 🤖
Google DeepMind's CEO calls for regulation before frontier AI rollout spirals.

Demis Hassabis has published a detailed proposal for an independent, U.S.-based organization — modeled on financial self-regulator FINRA — that would safety-test frontier AI models before public release. Under the framework, labs developing frontier-class systems would voluntarily submit them for review 30 days before launch, with screeners evaluating risks such as deception, bioweapons creation, and malicious hacking. Hassabis warned that open-source capabilities could enter dangerous territory within 18 months and urged the body to be operational this year. The plan also includes provisions for coordinating a slowdown among frontier labs if reviewers deem it necessary — a major concession that could reshape how the most powerful models reach the public. This is the most concrete governance proposal yet from an AI industry leader, directly pressuring competitors to respond.

🎓 Free Claude access for K-12 teachers 🎓
Anthropic removes the pricing barrier to embed Claude in U.S. classrooms.

Anthropic has officially launched Claude for Teachers, giving verified U.S. K-12 educators a full year of free premium access to Claude. The tool is specifically built to support lesson planning aligned to state academic standards, making it one of the first frontier-lab AI products offered at no cost to teachers. By targeting educators directly and eliminating the cost barrier, Anthropic aims to embed its AI into everyday teaching workflows at scale — a distribution play that could lock in a generation of institutional users. The move sets a precedent among leading AI companies for how they approach the education sector, as competition for institutional adoption intensifies among frontier labs.

💰 DeepSeek eyes $1.5B raise at $71B valuation 💰
China's efficient-AI challenger seeks capital for expansion ahead of a potential IPO.

Chinese AI lab DeepSeek is reportedly in discussions to raise approximately $1.5 billion at a $71 billion valuation, ahead of a potential initial public offering targeted for late 2026 or 2027. If confirmed, this would mark one of the largest private funding rounds for a Chinese AI company. DeepSeek has gained global attention for its efficient open-weight models that rival Western frontier labs at a fraction of the compute cost, and a public listing would give it access to deeper capital markets. The fundraising signals continued investor appetite for Chinese AI despite U.S. export controls and broader regulatory uncertainties — competitive pressure Western labs can no longer ignore.

INTERESTING TO KNOW

📱 Gemma 4 runs natively on Pixel 10 TPU 📱

Google has unveiled Gemma 4 E2B for TPU, a model variant built for native integration with the Pixel 10's on-device tensor processing unit. The rollout enables instant offline conversations, image identification, audio transcription, and system-level phone commands — all without a cloud connection. Every interaction stays private and local, demonstrating Google's strategy of tightly coupling its open Gemma model family with its own silicon — a capability jump that turns the Pixel 10 into a fully autonomous AI device.

🔊 OpenAI's Jony Ive AI speaker targets 2027 launch 🔊

OpenAI's first hardware device will be a screen-free, battery-powered AI speaker designed by Jony Ive, with a potential 2027 rollout according to Bloomberg. Powered by GPT-Live, the portable speaker features onboard cameras and sensors, handles questions, music, and smart-home control, and incorporates mechanical movement designed to make it feel "alive." A key differentiator is deep personalization — the device will study user data such as emails for tailored responses. However, Apple's trade-secrets lawsuit now looms over the project, adding legal risk that could reshape the launch timeline.