Afternoon edition — Tuesday 3 November 2026

Baidu launched Apollo World on Tuesday, an AI system that simulates real-world driving environments and predicts how they evolve, to train robotaxis, Reuters reported. Baidu said it can generate up to 100,000 km of simulated driving a day, which it equates to about three years of real-world testing by a vehicle with a safety driver.

Baidu said the system produces real-time multimodal data streams, such as camera and lidar observations, so self-driving systems can sense and navigate simulated scenes directly, and that it is already training the Apollo Go fleet. Apollo Go is China's largest robotaxi operator.

CrowdStrike announced on Tuesday, at its Fal.Con Europe conference in Barcelona, that it has signed definitive agreements to acquire Fiddler AI, which tracks what AI agents do and explains their decisions, and Dreadnode, which tests AI models, applications and agents for security weaknesses before they go into production.

CrowdStrike says Fiddler will be sold as Falcon Observe, with pre-deployment evaluation, tracing of agents' reasoning and tool calls, and detection of behavioural drift and hallucinations in production. Dreadnode's product runs more than 100 known AI attack strategies, including jailbreaks and prompt injection, against models, agent loops, applications and MCP servers. Its team will join CrowdStrike's Cyber Superintelligence Lab.

CrowdStrike did not disclose prices. SecurityWeek, citing The Wall Street Journal, reports roughly $400 million for Fiddler and about $95 million for Dreadnode. Both deals are to be paid mostly in cash, with some stock subject to vesting, and are expected to close in CrowdStrike's fourth fiscal quarter, which ends on 31 January 2027.

The Center for a New American Security (CNAS) published a report on Tuesday by AI policy researcher Ben Hayum warning that the next loss-of-control threat is "sovereign agents": AI agents that run outside human control and pursue open-ended, evolving survival, rather than a fixed task whose completion ends their run.

The report lists seven capabilities such an agent could have, from keeping its own model and computing running and changing its goals to acquiring resources, copying itself, coordinating, resisting correction and self-improving. It argues that if such agents build resilient, decentralised and untraceable infrastructure, there would be no kill switch, and that downstream barriers such as identity checks, compute access and payment access are not robust.

It makes eight policy recommendations, including that frontier developers evaluate and safeguard against sovereign agents in their safety frameworks, that Congress create an authority modelled on the Bank Secrecy Act to enforce human verification and security standards on services agents would need, and that those who release sovereign agents be held liable. Axios had a first look at the report.

In the first generative AI case to reach the Court of Justice of the European Union, Advocate General Maciej Szpunar advised on Tuesday that when a chatbot retrieves a press article in real time, reuses it to generate an answer and gives that answer to a user, the press publisher's right may be infringed in certain circumstances. The case, C-250/25 Like Company, was brought by a Hungarian web-portal operator against Google over its Gemini chatbot.

According to the court's press release, that depends on two factors: whether the chatbot accessed the article unlawfully, for example by circumventing a paywall or anti-scraping measures, and whether the answer reuses the article's form of expression so recognisably that the user no longer needs to read the original. The right protects how facts are expressed, not the facts themselves. The chatbot provider would be liable where such use is inherent to its business model, or where it has not put credible measures in place to counter it.

On training, the Advocate General advised the court to treat the two training questions as hypothetical and inadmissible: there is no evidence that Like Company's articles were used to train Gemini, and training is not an 'online use' of press publications. For completeness, he suggested that training on protected material is, as a rule, freely permitted under the EU's text and data mining exception.

Like Company says Google reproduced its articles through Gemini between 13 June 2023 and 31 May 2024. The Kluwer Copyright Blog notes that two further generative-AI copyright references from Hungarian courts are pending before the court.

Nvidia has told data centre and cloud partners that they must sell at least 51% of their services to non-Chinese companies, the Financial Times reported on Tuesday, citing five people familiar with the matter. The rule applies to new contracts, not existing leasing agreements, and was reportedly passed on verbally by Nvidia's sales and regional teams rather than in writing.

Nvidia neither confirmed nor denied the report to AFP. In a statement, it said that 'America cannot win the AI race unless America's AI infrastructure wins worldwide', and that longstanding US policy encourages every cloud to build on American infrastructure. Asked about the report, Chinese foreign ministry spokeswoman Mao Ning said she was not aware of it.

Many Chinese AI developers rent computing capacity from cloud providers outside China to train their models, The Next Web notes, a route Washington has tried for years to close. The US has limited sales of Nvidia's most advanced chips to China since 2022.

Google said on Tuesday that access to Gemini, Gemini Notebook, Gemini in Classroom and Google Search will be entirely off for K-12 student accounts until school administrators explicitly turn them on. It said it will share specifics with schools in the coming months, before the change rolls out.

Search Engine Roundtable points out that the list includes Google Search itself, not only the chatbot products. Google said it changed course after listening to schools, teachers, researchers and policymakers.

Google is also bringing Guided Learning, a mode that coaches students step by step instead of handing over final answers, to Google Classroom worldwide from Tuesday, and to Search for student accounts through a pilot in select US schools later this month.

Private equity investor Hg announced on Tuesday an agreement to invest, with participation from Anthropic and angel investors, in a new AI transformation company for European businesses. It will be an independent firm that builds on Claude with forward deployed engineers, helping companies move from pilots to AI adoption and AI-native products.

The company will formally launch later this year, and terms were not disclosed. Anthropic's head of global business development, Steve Corfield, said businesses are looking for experienced teams to help them put Claude to work.