THE RUNDOWN

Meta's new AI productivity assistant, Muse, has shot to the No.2 spot on US app charts within days of launch. It's the company's first serious push into AI-powered work tools — think task management, content creation, and what Meta is calling an 'AI agent' experience. Early users are impressed by the speed, but more than a few reviewers have noted it feels oddly perceptive in ways that are hard to shake. Whether that's clever engineering or something worth scrutinising, it's clearly connecting with people fast.

In a detailed new report, Anthropic says it has documented persistent 'distillation attacks' — where AI companies essentially feed Claude's outputs back into their own models to copy its capabilities without building them from scratch. The alleged culprits? Alibaba, Moonshot AI, and DeepSeek, all China-based players. Anthropic also separately confirmed that researchers have tried to use Claude to advance biological weapons work, which the company flagged in a broader misuse case study. This kind of transparency is rare, and it puts a spotlight on the security and ethics challenges sitting right at the heart of the AI race.

Nvidia CEO Jensen Huang made the case this week that his company's extraordinary growth isn't slowing down — he's projecting around 70% revenue growth in the coming year. He pushed back on concerns that Nvidia's deals are circular or self-serving, pointing to real demand across data centres, autonomous systems, and AI infrastructure worldwide. When you've got your chips in practically every major AI project on the planet, that kind of confidence isn't entirely surprising.

OpenAI has temporarily paused new Pro subscriptions, blaming demand from its Astra product for straining its systems beyond capacity. It's a strange position for the world's most talked-about AI company — basically telling customers 'we're too popular right now.' They say more capacity is coming, but it's a reminder that even the biggest players in the AI space are still figuring out how to scale.

A new US federal programme kicked off in Texas today, putting electric, hybrid-electric, and autonomous aircraft through real-world testing before the regulatory framework to govern them even exists. It's a deliberately ambitious move — test first, write the rules later. Air taxis have been 'almost here' for years, but actual government-backed test flights on American soil mark a meaningful step forward for urban air mobility.

Universal Music Group is partnering with ElevenLabs to launch an AI platform that lets users create remixes, mashups, and original compositions using UMG's licensed catalogue. It's a significant shift for a label group that spent years suing AI companies for using its music without permission. The move suggests the majors have decided that if AI music is inevitable, they'd rather control the pipeline — and take a cut.

Slack is rolling out a feature that lets users build interactive charts, dashboards, polls, presentations, and even mini-websites directly inside a chat window. It's the kind of thing that blurs the line between a messaging tool and a full productivity suite. For teams already living in Slack, it could genuinely reduce the number of apps they need open at once — and it shows how much the 'chat as workspace' idea has matured.

German fusion startup Proxima Fusion is building a €140 million factory to manufacture high-temperature superconducting tape — a critical material for fusion reactors that's currently dominated by Asian suppliers. It's a strategic bet on energy independence as much as it's a fusion play, and it signals that Europe is serious about owning its place in the clean energy supply chain rather than outsourcing a vital component.

California's governor has signed what's being called landmark legislation targeting how social media platforms and AI chatbots interact with minors. The laws aim to restrict certain features and data practices for users under 18, though they're not without critics who argue the rules are either too vague or don't go far enough. It's the latest signal that US states aren't waiting for federal action when it comes to tech regulation and youth safety.

Identity verification platform IDScan is now offering free credit monitoring after it was linked to a leaked database of driver's licences connected to an FBI investigation into a dark web marketplace. If you've ever used an app or service that required you to upload your ID for verification, this is a good reminder of how exposed that data can be — and how little you know about where it ends up. Data security for identity documents remains dangerously underregulated.

That's your September 11 Rundown. The AI story is moving fast — but the security story underneath it is moving just as fast, and it's the one that tends to matter more when things go wrong.