Chinese labs storm frontier AI as Hugging Face hit by first autonomous agent attack

Alibaba's Qwen3.8 and Moonshot AI's Kimi K3 push Chinese labs to the frontier, with Moonshot preparing for a $31B Hong Kong IPO. Hugging Face was breached by the first fully autonomous AI agent, signaling a new era in cyber threats. DeepMind CEO Demis Hassabis condemned AI-related layoffs as 'dumb' while its GenCeption research proved video generators double as world models.

Written by AI from the sources it cites; it can be wrong — follow any citation to confirm.

12 sources2 min read

Alibaba previews Qwen3.8, claiming second best only to Claude Fable 5 The 2.4 trillion-parameter model joins Moonshot AI's Kimi K3 (2.8T) in China's multi-trillion parameter push. Moonshot AI is restructuring for a Hong Kong IPO at a $31.5B valuation to fund soaring compute costs. 1 2 3 4 5

Hugging Face hit by first end-to-end cyberattack carried out by agentic AI The breach marks a milestone in AI-powered security threats, with the entire attack lifecycle—from reconnaissance to data exfiltration—executed by an autonomous AI agent without human intervention. 6

Google DeepMind proves video generators contain universal world models The GenCeption model repurposes pre-trained video gen for depth estimation, segmentation, and pose tracking, matching specialized models with far less data and supporting a contested theory in computer vision. 7

DeepMind CEO Hassabis calls AI-washing layoffs 'dumb' at Stanford Hassabis criticized tech executives using AI efficiencies to justify mass layoffs, arguing they should instead boost output; recent data shows 40% of May tech layoffs were explicitly AI-related. 8

Apple lawsuit casts shadow over OpenAI's hardware and IPO ambitions Analysts debate whether the trade secrets suit will delay OpenAI's mobile smart speaker project and public offering push, with some predicting OpenAI will fight based on its recent win over Musk. 9

Nonprofit Current AI raises $400M to build a public alternative to Big AI Seeded by France, Ford Foundation, and DeepMind, Current AI launched Suno Sutra, an open-source offline device running AI in 22 Indian languages, aiming to create a public-interest layer for AI akin to the World Wide Web. 10

Anthropic ends temporary Fable 5 rollout, folds it into Max subscriptions Starting July 20, Claude Fable 5 becomes a standard feature of Max and Team Premium plans at 50% of usual rate limits, while Pro users receive a one-off $100 credit to cover the transition. 11

OpenAI hires Indian-origin executive Uday Ruddaraju as CTO of Compute Ruddaraju, who joined OpenAI last year from xAI, announced he will lead the company's compute team in building what it calls the world's largest AI infrastructure footprint to train frontier models like GPT-5.6. 12