AI News Briefing — July 5, 2026

Following up on yesterday's news that Anthropic plans to develop its own drugs, Anthropic is launching its own drug development programs targeting neglected diseases that pharmaceutical companies deem unprofitable. Novartis CEO Vas Narasimhan estimates AI could shorten drug development from twelve years to seven or eight and roughly double the success rate.

A study of more than 26,000 Chinese students found AI users completed homework faster and scored higher initially but performed up to 24 percent worse on exams. The negative learning effects took about two years to fully surface, suggesting short-term studies underestimate the harm.

NVIDIA Research introduced HORIZON, a hands-free agent framework that treats hardware design as repository-level code evolution, evolving isolated git worktrees and committing only when acceptance gates pass. It reports 100 percent completion across evaluated RTL benchmark suites while acknowledging agentic hardware design is not yet solved.

Anthropic released Claude Science in beta, a multi-agent AI workbench for reproducible research pipelines spanning genomics, proteomics, and cheminformatics. It runs on existing Claude models rather than a new one, integrating research tools, executing multi-step analyses, and recording provenance for Pro, Max, Team, and Enterprise users.

NVIDIA and academic collaborators introduced ASPIRE, a self-improving robotics framework using code-as-policy with fine-grained failure feedback and retained fixes across tasks. It reaches 31 percent zero-shot performance on long-horizon LIBERO-Pro tasks, addressing the limitation of agents discarding learned solutions.

As first reported in yesterday's News coverage of Claude Code's China problem, Alibaba has reportedly classified Anthropic's Claude Code as high-risk software and banned employees from using it. The move signals corporate security concerns around external AI coding tools, particularly across US-China boundaries.

A Guardian investigation reports that OpenAI's much-touted Stargate UK datacenter project, framed by ministers as up to 30 billion pounds of investment, appears to have been largely hypothetical, with 20 billion pounds unbacked by firm commitments. The plans were paused in April amid concerns over UK regulation and high energy costs.

The NHS in England will deploy AI within its app to triage patients and direct them to appropriate services such as GP appointments, pharmacies, or A&E, as part of a 10 billion pound systems overhaul. The rollout is expected to reach roughly 200,000 patients over the next year.

In its ongoing copyright litigation with Disney, Universal, and Warner Bros, Midjourney is seeking to compel the studios to disclose how they themselves use AI. The discovery move aims to reframe the legal dispute by highlighting the studios' own generative-AI practices.