AI News for July 4, 2026 — 7min.ai

[NEWS] Anthropic's life sciences head says it will develop drugs for neglected diseases: Speaking at an event called The Briefing: AI for Science, Anthropic's head of life sciences, Eric Kauderer-Abrams, said the company will go further than its existing drug discovery program and develop its own drugs, focusing on treatments for neglected diseases.
[NEWS] Alibaba bans staff from Claude Code as Anthropic moves to block Chinese access: Alibaba banned employees from using Anthropic's Claude Code and ordered them to delete all Claude models, adding the tool to a list of high-risk software after reports of hidden code that flagged users based in China.
[NEWS] Kling AI's HKEX filing confirms $2.04B raised at $18B valuation, Kuaishou stake falls to 68%: Kuaishou's AI video division Kling AI confirmed via a Hong Kong Exchange filing that it has raised roughly $2.04 billion (13.82 billion yuan) from investors including CPE, Guofang Investment, BlueFive, Tencent, and Citic Securities, valuing the unit at $18 billion.
[RESEARCH] ByteDance finds AI agents double their learning speed every three months: ByteDance's Seed AI team published a new scaling law showing that AI agents can double how fast they improve by executing real-world tasks, with the doubling happening roughly every three months.
[RESEARCH] Bridgewater and Thinking Machines fine-tune an open model that beats GPT and Claude on finance: Bridgewater's AIA Labs and Mira Murati's Thinking Machines Lab fine-tuned Qwen3-235B on internal expert judgments and beat leading commercial models at evaluating financial documents, hitting nearly 85% accuracy at 14 times lower cost.
[RESEARCH] Mistral releases Leanstral 1.5, an open-source model that aces formal math and catches real bugs: Mistral released Leanstral 1.5, a free Apache 2.0 model built for formal verification in the Lean 4 language.
[RESEARCH] Alibaba's Elements Claw AI agent discovers four new superconductors: Alibaba's Damo Academy unveiled what it calls the first AI agent built to discover superconducting materials, and says the tool has already found four previously unknown compounds that were later verified in lab experiments.
[RESEARCH] Security vulnerability reports spike 3.5x as AI models start hunting bugs: Epoch AI charted a sharp jump in disclosed software flaws: in June 2026, organizations reported roughly 1,500 high-severity and critical CVEs, more than 3.5 times the previous monthly record.
[NEWS] Google DeepMind and A24 form a first-of-its-kind research partnership: Google DeepMind and film studio A24 announced a multi-project research collaboration to develop new tools and workflows for filmmakers, and Google made an investment in A24.
[TOOLS] Microsoft plans a Copilot super app with background AutoPilot agents: Microsoft will overhaul Copilot in August, merging its consumer and enterprise apps into one product with AI coding tools and new background agents called AutoPilot that handle scheduling and email summaries.

About 7min.ai · Subscribe to newsletter