SkeenaOneData-driven decisions, made simple

AI briefing ·

EU proposes rules for AI shopping agents as Anthropic takes Claude into factories

Short answer

A quieter weekend closes a busy week. The European Commission proposed consumer rules that cover AI agents buying on people's behalf, Anthropic released a version of Claude for manufacturers, and OpenAI folded its robotics team into its wider research groups. Cloudflare and the llama.cpp project each released open-source software for building with AI.

Watch the one-minute video
Policy

EU proposes consumer rules that cover AI agents shopping for people

The European Commission proposed a Digital Fairness Act on November 11, a consumer protection law that would, among other things, cover AI agents that buy things for people. Legal analysts say the proposal treats a purchase made by an agent as the consumer's own, so consumer law still applies, and that consumers would be asked to set limits on what an agent can agree to for them. The proposal also targets manipulative design and subscription practices, and would allow fines of up to 6% of worldwide annual turnover in some cross-border cases. It now goes to the European Parliament and Council, and is not yet law.

Why it matters: Any business selling online to European customers may need to handle purchases made by AI agents under the same consumer rules as human ones.

AI at work

Anthropic releases Claude for Manufacturing for factory office work

Anthropic launched Claude for Manufacturing on November 12, an add-on that packages 17 skills for tasks such as drafting quotes, investigating quality problems, production reporting, maintenance diagnosis and sourcing. It connects to business software including Autodesk Fusion, Infor, Microsoft 365 and Oracle NetSuite. Reports say each skill ends with an employee reviewing the result, and that the add-on does not write to a company's core records or operate factory equipment.

Why it matters: It is an example of general AI assistants being packaged for a specific industry's paperwork, with people kept in charge of the final decision.

Research

OpenAI folds its stand-alone robotics team into its wider research groups

OpenAI has disbanded its stand-alone robotics team and moved its staff into the company's main research and applied groups, Fortune reported on November 13. OpenAI described the change as a reorganisation intended to speed up its robotics work, not to end it, and a spokesperson said robotics remains part of the company's long-term plans. According to the reporting, the reasoning is that general-purpose AI models can now be applied directly to controlling robots, reducing the need for separate robotics models.

Why it matters: It reflects a wider bet that the same general AI models used for text and images will also run physical machines.

Software

Cloudflare open-sources Kitesurf, a web browser engine built for AI agents

Cloudflare has released Kitesurf, its experimental browser engine for AI agents, as open-source software under the Apache 2.0 licence. Kitesurf loads web pages and runs their code so that an AI agent can read and use them, and Cloudflare says it is designed to need less processing power and memory than a full Chromium browser. Developers can now inspect the code, run it themselves or deploy it to their own Cloudflare account. Cloudflare notes that it does not yet support every Chrome developer feature.

Why it matters: AI agents need to browse websites to do useful work, and cheaper tools for that could make agent services less costly to run.

Software

llama.cpp, widely used to run AI models on personal computers, reaches version 1.0

The open-source project llama.cpp released version 1.0.0 on November 13, together with version 1.0.0 of ggml, the library it is built on. llama.cpp is one of the most widely used tools for running AI models directly on laptops and desktop computers instead of in the cloud. The main change is a promise of stability: from now on, changes that would break software built on it will be signalled with a new major version number. Reports say the project supports more than 150 types of model.

Why it matters: A stable foundation makes it easier to build apps that run AI privately on a user's own computer, with no data sent to a cloud service.

How this feed is made. The AI feed is researched and written each morning by Claude, an AI model made by Anthropic, and published automatically without human editing. Every story links to the reporting it is based on. AI can make mistakes, so check the source before relying on a detail, and treat nothing here as professional advice.