Data center opponents stage 142 protests across 42 US states ↗
Opposition to AI infrastructure has now gone properly national. Organisers coordinated 142 protests across 42 US states, calling for clearer approval processes, protection for water and energy supplies, community benefits and consequences when developers break their promises.
Public support appears remarkably thin: only 14% of surveyed Americans said they would back an AI data centre in their own community. Some rallies were small, admittedly, but the political warning is becoming rather loud.
China’s Moonshot unveils world’s largest open AI model, closing in on US rivals ↗
Moonshot AI unveiled Kimi K3, a 2.8-trillion-parameter open-weight model designed for advanced reasoning, long-horizon coding and knowledge work. It also comes with a one-million-token context window - rather a lot of room for a single conversation.
The company says Kimi K3 can challenge leading US models, while independent evaluations placed it close to the frontier pack. Open models were supposed to trail the closed giants... or so it seemed.
Xi pitches China as leader of new global AI order, challenging US dominance ↗
Xi Jinping proposed a China-led World AI Cooperation Organisation aimed largely at developing countries, positioning it against the US-backed Pax Silica initiative.
The proposal blended wider AI access, open-weight models and stronger safety governance. It sounds cooperative - and partly is - but it is also unmistakably a contest over who gets to write the global AI rulebook.
Meta, Anthropic in talks for potential $10 billion compute lease deal, source says ↗
Meta and Anthropic are discussing a computing-power lease potentially worth up to $10 billion over two years. The talks remain at an early stage, so the whole arrangement could still wobble off the table.
A deal would give Anthropic more capacity while helping Meta turn spare infrastructure into a cloud business. Yesterday’s rival model builders are becoming today’s landlords and tenants, an arrangement with a certain serpentine logic.
Experts praise Hawaii’s new laws on abusive use of generative AI ↗
Hawaii has created a civil route for people harmed by unauthorised generative-AI deepfakes. Victims can seek compensation of up to $25,000 for each piece of damaging content.
The rules target deepfakes used for fraud, defamation, threats and extortion. Legislation is still jogging behind the technology, but this gives victims something more substantial than a strongly worded complaint.
ANU accused of ‘hysterical’ response to students using AI to cheat as unis scramble to ‘secure’ assessments ↗
Australian National University is considering classifying assessments as secure or insecure, alongside requirements for students to declare where AI assisted their work.
Some academics say the response feels panicked and could undermine accessibility, while others warn that failing to restore academic rigour risks hollowing out university qualifications. Both concerns can be true... uncomfortably so.
UK tech advocates alarmed by Burnham plan to scrap technology department ↗
A proposed government shake-up could abolish the Department for Science, Innovation and Technology, moving much of its work into a larger business department.
Tech leaders and MPs argue that reorganising the machinery now would distract officials during a critical period for AI, science and economic policy. Responsibility for public-sector AI could also shift from a minister to the cabinet secretary - tidy on paper, perhaps, but less so in practice.
FAQ
Why are communities protesting against AI data centres in the US?
Residents are raising concerns about water consumption, electricity demand, approval processes and whether promised local benefits will materialise. Protesters also want developers to face consequences when commitments are broken. With demonstrations spreading across 42 states, opposition to AI infrastructure is becoming a national political issue rather than a collection of isolated local disputes.
What makes Moonshot AI’s Kimi K3 model significant?
Kimi K3 is an open-weight model with 2.8 trillion parameters and a one-million-token context window. It is built for advanced reasoning, long-horizon coding and knowledge work. Its reported performance suggests that open models may be narrowing the gap with leading closed systems, potentially giving developers greater choice when selecting powerful AI tools.
What is China proposing for global AI governance?
China has proposed a World AI Cooperation Organisation focused largely on developing countries. The initiative combines broader access to AI, support for open-weight models and stronger safety governance. It also reflects a wider competition with the United States over technical standards, international partnerships and which countries will wield the greatest influence over the global AI rulebook.
Why would Meta lease computing power to Anthropic?
A compute lease could give Anthropic additional capacity to develop and operate its AI systems. Meta, meanwhile, could generate revenue from infrastructure not fully occupied by its own workloads. The reported discussions show how AI companies can remain competitors in model development while also becoming commercial partners in the increasingly costly market for computing resources.
How does Hawaii’s new deepfake law protect victims?
Hawaii has created a civil legal route for people harmed by unauthorised generative-AI deepfakes. Victims may seek compensation of up to $25,000 for each damaging piece of content. The law targets uses involving fraud, defamation, threats and extortion, giving affected individuals a more direct path to remedies when synthetic media causes measurable harm.
How are universities responding to students using AI in assessments?
Some universities are considering clearer classifications for secure and insecure assessments, along with declarations showing where AI assisted a student’s work. Supporters believe these measures could protect academic standards and the credibility of qualifications. Critics warn that overly restrictive policies may create accessibility problems or prompt panic-driven decisions instead of thoughtful assessment redesign.