US says it will speed development and use of AI for national security ↗
The White House moved to accelerate AI use across intelligence and military work, while saying the tech should not be used for unlawful surveillance or censorship.
The memo also pushes the Pentagon to update rules around autonomous weapons. So, yes - faster AI adoption, but with a very loud “please don’t let this become a sci-fi drawer fire” attached.
China bets on AI to promote President Xi Jinping's thinking ↗
Xinhuanet, owned by China’s official Xinhua news agency, plans to spend over 1.1 billion yuan on an AI agent built around Xi Jinping’s political ideology.
The tool, called Xinhua Yudian, is pitched as an “authoritative” assistant for studying, citing, and spreading official doctrine. Unnervingly tidy, and also a little ominous - like a chatbot wearing a party badge.
Google will pay SpaceX $920M per month for compute ↗
Google is set to pay SpaceX $920 million per month for access to AI compute, including roughly 110,000 NVIDIA GPUs plus supporting hardware.
The explanation: demand for Google’s agent platform has run hotter than expected. The AI boom is starting to look less like software and more like a global power-grid eating contest.
AirTrunk commits $30B to build 5GW of AI data centers in India ↗
Blackstone-backed AirTrunk said it will invest $30 billion to build 5GW of data center capacity in India.
The move puts India deeper into the AI infrastructure race, with cloud and compute demand pulling huge capital into new regions. Not glamorous, maybe, but this is the plumbing under the magic trick.
The token bill comes due: Inside the industry scramble to manage AI’s runaway costs ↗
Companies are now feeling the cost of all those “AI everywhere” promises. Uber reportedly burned through its AI coding budget early, Microsoft pulled some Claude Code licenses, and Cursor renewals are getting pricier.
The twist is slightly brutal: per-token prices can fall while total spending still explodes. More agents, more usage, more hidden invoices - the spreadsheet goblin always wins.
Exclusive: Dems introduce bill for "responsible" defense AI ↗
U.S. senators Chris Coons and Jack Reed are introducing a bill to set rules for how the Pentagon buys, tests, and uses AI-enabled autonomous weapons.
The proposal focuses on human oversight, manual override capability, privacy checks, and limits on nuclear decision-making. That sounds basic, but in defense AI, “basic” is doing a terrifying amount of work.
Labour will make AI ‘work for the workers’, says Liz Kendall ↗
UK technology secretary Liz Kendall said Labour wants AI to enhance work rather than simply sweep jobs away.
She pointed to skills schemes, disadvantaged schools, and support for young people at risk of falling out of education or work. It’s policy with a hard hat on - maybe sturdy, maybe optimistic, but definitely aware the floor is moving.
FAQ
What is changing in US AI national security policy?
The White House wants to accelerate AI development and adoption across intelligence and military work. At the same time, the memo says AI should not be used for unlawful surveillance or censorship. It also pushes the Pentagon to update rules around autonomous weapons, suggesting that faster deployment is being paired with clearer limits.
How is China using AI to promote Xi Jinping’s political ideology?
Xinhuanet, which is owned by China’s official Xinhua news agency, plans to spend more than 1.1 billion yuan on an AI agent based on Xi Jinping’s thinking. The tool, called Xinhua Yudian, is being positioned as an authoritative assistant for studying, citing, and spreading official doctrine. The project shows how AI can be used for political messaging, not just productivity.
Why would Google pay SpaceX for AI compute?
Google is reportedly paying SpaceX for access to major AI compute resources, including around 110,000 NVIDIA GPUs and related hardware. The stated reason is stronger-than-expected demand for Google’s agent platform. This points to a wider trend: advanced AI products increasingly depend on vast physical infrastructure, not just software improvements.
What does AirTrunk’s India data center investment mean for AI infrastructure?
AirTrunk’s planned $30 billion investment in India would add 5GW of data center capacity. That matters because AI systems need large amounts of cloud infrastructure, electricity, cooling, and compute hardware. The move also suggests India is becoming a more important region in the global AI infrastructure race as demand spreads beyond existing hubs.
Why are AI costs rising even when token prices fall?
AI costs can still rise because total usage grows faster than per-token prices decline. More teams are adding AI tools, agents are running more tasks, and usage can become harder to track across large organizations. The article points to companies facing budget pressure from coding tools and AI subscriptions, showing how quickly small unit costs can swell into large bills.
What rules are being proposed for defense AI and autonomous weapons?
U.S. senators Chris Coons and Jack Reed are introducing a bill focused on responsible defense AI. The proposal covers how the Pentagon buys, tests, and uses AI-enabled autonomous weapons. Key ideas include human oversight, manual override capability, privacy checks, and limits on nuclear decision-making. The goal is to create safeguards before military AI becomes more widely deployed.