AI Daily Brief: 5 July 2026

5 July 2026

Quick Read: NHS England will put AI triage into the NHS app for 200,000 patients over the next year after a Sussex trial cut phone queues by 29%. Alibaba is reportedly banning Claude Code from 10 July as Anthropic tightens access controls linked to China. Security researchers say attested TLS flaws could undermine confidential computing, with CVE-2026-33697 rated 7.5. Kuaishou's Kling AI has drawn a $2.8bn funding round at a $15bn pre-money valuation.

Today is a practical AI brief rather than a launch spectacle. The biggest stories are about public service triage, enterprise controls, model access risk, creative copyright discovery and the security assumptions behind sovereign cloud.

NHS app AI triage rollout will reach 200,000 patients in the next year

NHS England says the NHS app will use AI to ask patients questions and direct them to a GP appointment, pharmacy, A&E, community service or self-care advice. The first phase is expected to reach more than 200,000 patients over the next 12 months, with full availability for app users planned by April 2028.

The strongest early number is a Sussex trial at Wealden Ridge Medical Partnership, where the tool was linked to a 29% reduction in people queuing by phone for appointments. The wider programme sits inside a £10bn technology, digital and data investment, including AI notetaking pilots across NHS trusts.

Our take: For UK organisations, this is the useful AI pattern to study: a narrow workflow, clear escalation points and measurable operational pressure. The risk is not the tool itself. The risk is rolling it out faster than patient safety, privacy, digital exclusion and clinical accountability can be governed.

Alibaba reportedly bans Claude Code after China access controls row

Alibaba will reportedly ban employees from using Anthropic's Claude Code from 10 July, with the tool classified as high-risk software. The move follows reports that Anthropic has been closing loopholes that allowed China-linked users to access Claude despite restrictions on Chinese companies and foreign entities owned by them.

TechCrunch reports that Anthropic described one controversial Claude Code mechanism as an experiment intended to prevent account abuse by unauthorised resellers and protect against distillation. Alibaba is reportedly directing employees to use its own Qoder tool instead.

Our take: This is a live example of AI supply chain risk moving from abstract policy to developer tooling. Businesses using AI coding assistants need to know where prompts, repositories and telemetry go, and whether a vendor can change access rules, detection methods or compliance behaviour without much operational notice.

Confidential computing research challenges a core sovereign cloud assumption

The Register reports on formal research into attested TLS, the protocol layer used in confidential computing to prove that a server is running inside a genuine, unmodified trusted execution environment. Researchers from TU Dresden and collaborators found diversion and relay attacks where clients can verify trustworthy evidence but still send traffic to the wrong machine.

The issue has been assigned CVE-2026-33697 with a 7.5 high-severity rating, and the work analysed real implementations including Meta's Private Processing system for WhatsApp, Edgeless Systems' Contrast and Cocos AI. The researchers argue that current intra-handshake attestation may not be able to prove the strongest level of binding without changing assumptions inside TLS 1.3.

Our take: Sovereign cloud is often sold as a procurement answer, but this story shows why architecture matters. If an organisation is relying on confidential computing for regulated AI workloads, it should be asking for evidence about attestation design, threat models, patches and independent verification, not just a sovereignty label.

Midjourney wants Hollywood studios to disclose their own AI use

In its copyright dispute with Disney, Universal and Warner Bros., Midjourney is asking the court to force the studios to disclose more detail about their own use of generative AI. The studios sued Midjourney over alleged infringement after its models generated images of protected characters including Bart Simpson, Darth Vader, Superman and Batman.

A judge had already required some disclosure, but only for AI use that led to consumer-facing images or videos. Midjourney argues that the studios should also disclose internal AI work, including storyboarding and ideation, because that may support its fair use defence and market harm arguments.

Our take: The copyright fight is moving beyond whether models were trained on protected work. It is becoming a fight over industry custom. If major rights holders are using similar AI methods internally, it complicates simple public messaging about acceptable and unacceptable AI use.

NCA and IWF warn parents about public child images and AI abuse risks

The National Crime Agency and Internet Watch Foundation have warned parents not to publicly post images of children because of the growth of AI-generated child sexual abuse material. The IWF said it identified more than 8,000 AI-generated realistic child abuse images and videos in 2025, a 14% increase on the previous year.

The video trend is particularly stark. The IWF said analysts found 13 AI-generated child abuse videos in 2024, rising to 3,440 in 2025. The organisations are urging parents to review privacy settings and limit image sharing to trusted groups.

Our take: This is not just a consumer safety issue. Any organisation that holds images of children, staff or vulnerable people needs stronger data minimisation, permissions and takedown processes. AI abuse risk changes the value and sensitivity of ordinary image archives.

Kling AI raises $2.8bn as Chinese video models draw strategic capital

Kuaishou Technology's Kling AI has reportedly secured a funding round of more than 19bn yuan, about $2.8bn, from investors including Alibaba and Tencent. Reports put the pre-money valuation of the AI video business at about $15bn.

The deal signals that video generation remains one of the hottest strategic categories in AI, even as copyright, compute cost and model safety questions intensify. It also shows that China's major platforms are still willing to fund domestic AI infrastructure and media-generation capability at scale.

Our take: UK businesses should read this as a sign that AI video will keep improving and spreading into marketing, training, product demos and customer support. The governance work should start now: rights clearance, brand controls, disclosure rules and practical review workflows.

Yann LeCun backs world models as a route beyond current LLM limits

Former Meta chief AI scientist Yann LeCun told the BBC that current systems such as ChatGPT, Claude and Gemini are not a path to human-like or animal-like intelligence because they do not understand the physical world. His new Paris-based company, AMI Labs, is working on Joint Embedding Predictive Architecture, or JEPA, as an alternative approach.

AMI Labs has reportedly raised more than $1bn in seed funding from investors including Nvidia and the fund managing Jeff Bezos' private wealth. The broader world models field aims to build systems that reason about causes, actions and possible futures rather than only producing statistically plausible outputs.

Our take: The business point is not that LLMs are suddenly obsolete. It is that the next generation of AI products may split by use case: language models for knowledge work, world models for robotics, physical operations and simulation. Strategy should avoid treating one model class as the answer to every problem.

Quick Hits

Frequently Asked Questions

How often is the AI Daily Brief published?

Every morning at 7:30am UK time, covering the previous 24 hours of AI news from over 30 sources.

How are stories selected?

UK-relevant stories are prioritised first, then by business impact and practical implications for UK organisations adopting AI.

Why should business leaders follow AI news?

AI is moving faster than any technology in history. Staying informed is essential for making smart decisions about AI investment, adoption, and governance.