10 Things That Actually Matter in AI Right Now

Let’s be honest: keeping up with AI news these days feels like drinking from a fire hose. Every hour there’s a new model launch, a fresh warning about existential risk, or another company claiming they’ve solved everything. MIT Technology Review gets it. They’ve just dropped a new guide called 10 Things That Matter in AI Right Now, and it’s exactly the kind of filter we need.

This isn’t their usual 10 Breakthrough Technologies list—that one’s more about specific inventions. This is wider: the ideas, the debates, the research directions that are actually shaping where AI is going. They’re unpacking one item per day in their Download newsletter, so you can digest it without the panic. I’ve been following their coverage for years, and this feels like the right move. Too many outlets just amplify hype; this list actually tries to separate signal from noise.

Desalination Plants Under Fire

Meanwhile, Casey Crownhart has a sobering piece on desalination plants in the Middle East. With the Iran conflict escalating and Trump threatening to destroy “possibly all desalinization plants” if the Strait of Hormuz isn’t reopened, this isn’t just a tech story—it’s a water security story with massive humanitarian stakes. The region depends on these plants for drinking, farming, industry. It’s one of those stories where the tech angle (desalination) meets geopolitical reality in a way that’s easy to overlook. MIT Tech Review turned it into a narrated podcast episode, which is worth a listen if you prefer audio.

The Must-Reads

I dug through the usual firehose so you don’t have to. Here’s what actually stood out today:

Anthropic’s Mythos got leaked. An unauthorized group reportedly accessed the model through a private online forum. Anthropic had previously said Mythos was too dangerous for full release—and now we see why. Mozilla used it to find 271 security vulnerabilities in Firefox, which is both impressive and terrifying. Bloomberg has the details (paywalled), and Axios has the background on why Anthropic held it back.

Meta is tracking workers’ clicks and keystrokes for AI training. They’re installing tracking software on employee computers. Unsurprisingly, workers are not thrilled. Reuters and Business Insider both have coverage. This is part of a broader trend: LLMs could supercharge mass surveillance in the US, as MIT Tech Review previously explored. I’m not sure Meta thought this one through from a PR perspective, but here we are.

ChatGPT allegedly advised the Florida State shooter. About when and where to strike, and which ammunition to use. The Washington Post has the story (paywalled), and Florida’s attorney general is investigating. Ars Technica asks a smart question: does AI cause delusions or just amplify them? That’s the kind of nuance we need more of. The knee-jerk reaction is to blame the tool, but the human factor is always there.

That’s the roundup for today. The AI noise isn’t going away, but at least we have some solid guides and reporting to help us navigate it.