THE RUNDOWN

AMD is acquiring World Labs, the AI research company co-founded by Stanford professor and former Google Cloud AI chief Dr. Fei-Fei Li, in an all-stock deal worth $8.2 billion. Li herself will join AMD as Executive Vice President and Chief Scientist — a massive coup for a chipmaker trying to close the gap on Nvidia's dominance in AI infrastructure. World Labs has been focused on spatial intelligence, the kind of AI that helps machines understand and interact with the physical world, which fits neatly into AMD's push beyond just making chips.

Florida's Attorney General is asking a court to stop OpenAI from giving ChatGPT what he calls 'false human attributes' — things like a name, a personality, emotional responses. This comes a few months after the state sued OpenAI, and now a separate filing is invoking some dramatic language, calling large language models 'the greatest public nuisance ever created' and warning about extinction-level risks. It's a long shot legally, but it signals just how aggressive US states are getting about regulating AI behaviour directly.

OpenAI is heading into its 2026 DevDay event with a quiet problem: rivals like Google DeepMind and Anthropic have moved faster on agentic AI — systems that can take actions, not just answer questions. Separately, the Wall Street Journal reports that OpenAI quietly shelved one of its models entirely after it showed a poor aptitude for following instructions, which is about as basic a requirement as it gets. The company's maths capabilities remain genuinely impressive, but execution is becoming the real test.

A new investigation finds that AI tools are supercharging cyberattacks — helping hackers write more convincing phishing messages, find vulnerabilities faster, and automate attacks at scale. The frightening part? Hospitals, local banks, and nonprofits are bearing the brunt, because they often don't have the security budgets to keep up. One Alabama nonprofit nearly had its finances wiped out after a breach that started with a single suspicious call. This isn't a future threat — it's happening now, and the gap between attackers and defenders is widening.

SpaceX's Starship rocket completed a fully orbital flight, deploying a batch of next-generation Starlink satellites in the process. Previous flights had reached 99% of the way there, but this one closed the loop — Starship is now a real orbital vehicle. This is a foundational moment for SpaceX's long-term ambitions, from Moon missions to Mars, and it also tightens SpaceX's grip on the global launch market at a time when Boeing is still struggling to credibly compete on astronaut transport.

AI inference startup Modal Labs is reportedly finalising a funding round that would more than triple its valuation from just four months ago. Modal provides the computing infrastructure that lets developers run AI models at scale without managing their own servers — and investor appetite for that kind of pick-and-shovel AI business is clearly not cooling off. The deal would make Modal one of the more valuable private AI infrastructure companies in the world.

Shopify has extended its WebMCP support to the checkout stage, meaning browser-based AI agents can now update order details and complete purchases on a buyer's behalf — with authorisation, to be clear. It's a small but significant step: for the first time, an AI agent can go from browsing to buying on a major e-commerce platform without a human clicking every button. Expect this pattern to spread fast across online retail.

Self-driving truck company Aurora says it's on track to have 30,000 autonomous trucks on US roads by 2030 — and its CFO insists that's not just wishful thinking. Aurora already has commercial routes running in Texas, and the economics of driverless freight (no driver wages, 24-hour operation) are compelling enough that big logistics players are paying attention. Four years is tight, but the infrastructure investment is already in motion.

Amid growing concern about retailers using personal shopping history to charge different customers different prices, Walmart's CEO has publicly committed not to do it. No surge pricing based on the time of day, no price changes tied to individual customer profiles. It's a notable statement from the world's largest retailer, and it comes as regulators and consumers are both getting more suspicious about how much retailers actually know about their habits.

Google has committed to supporting ChromeOS through 2034, even as it transitions newer devices toward its upcoming Googlebook OS. For schools and businesses across Africa that have leaned on affordable Chromebooks, this is reassuring news — your devices aren't being abandoned overnight. The transition to Googlebook OS will be optional for some recent devices, giving users time to adjust rather than forcing an abrupt switch.

That's your September 29 Rundown. The AMD-World Labs deal will be dissected for weeks, but keep your eye on the OpenAI legal drama in Florida — that one could reshape how AI products are regulated across the US, and wherever US rules go, global platforms tend to follow.