Napster's AI teacher clones, Microsoft's internal scraping bombshell, and the FAA's $875M AI bet

There is a word for what today's newsletter is: contradictory. The same week AI lab CEOs are calling for a slowdown, a Microsoft exec privately called the industry's training data practices outright theft - while Microsoft was doing the exact same thing. Oh, and Napster is back. Yes, that Napster.

The answer is hiding near the bottom of today's newsletter... keep scrolling. 👇

Remember Napster? The platform that nearly destroyed the music industry in the early 2000s has apparently decided its second act involves digitally cloning classroom teachers with AI. According to Wired, the relaunched company is now targeting the education sector, building technology that can create AI versions of real instructors for use in schools.

The pitch is straightforward: take a real teacher's voice, mannerisms, and teaching style, and create a scalable AI replica that can interact with students outside of class hours. It's the kind of idea that sounds futuristic until you think about the consent, accuracy, and labor implications for actual educators for more than about 30 seconds.

If you're curious about how AI is reshaping classrooms more broadly, we've been tracking this at our AI Education tag page - the tension between genuine learning tools and replacement-by-automation keeps getting sharper.

Read the full story →

Whether AI teacher clones are a genuine accessibility tool or a cost-cutting shortcut dressed up in EdTech language depends entirely on who is deploying them and why - and that question deserves a real answer before this scales.

AI is moving fast and most people are falling behind. 

The Rundown AI keeps you ahead of the curve. 

It's a free AI newsletter that keeps you up-to-date on the latest AI news, and teaches you how to apply it in just 5 minutes a day.

Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses — tailored to your needs.

Sign up to start learning.

Here is the story that should be getting more attention. Newly unsealed court filings from the ongoing New York Times lawsuit against OpenAI and Microsoft reveal that a Microsoft executive privately described AI data scraping as 'the largest theft of labor in human history' - while Microsoft was actively doing the same thing.

According to TechCrunch, the filings show that both Microsoft and OpenAI scraped paywalled content from the Times, built training datasets from it, and executives internally warned this would gut publishers. The gap between private knowledge and public statements here is enormous.

The most striking detail is not just that someone inside Microsoft knew this was ethically fraught - it is that the company kept going anyway. This lands in the middle of a broader legal and regulatory debate about whether AI companies owe publishers anything for the content that trained their models.

We've been tracking the legal battles around AI copyright closely. If you want the full background on how we got here, our archive has the full timeline.

Read the full story →

When internal documents contradict public positions this dramatically, it tends to matter a lot in court - and this filing could shift the legal landscape for how AI companies defend their training data choices going forward.

The Federal Aviation Administration is launching a new AI-based software program costing $875 million, designed to help air traffic controllers better manage the increasingly complex flow of aircraft across US skies. According to TechCrunch, the system is intended to give controllers better decision support tools rather than replace human judgment entirely.

This is worth paying attention to beyond the headline number. The FAA has been operating on aging infrastructure for years, and the combination of rising drone traffic, increasing commercial flights, and staff shortages has pushed the system to its limits. AI as a co-pilot for human controllers - rather than a full replacement - is the framing here.

Unlike most AI deployments where a bad output means a wrong answer or a strange image, errors in air traffic control have catastrophic consequences. The fact that the FAA is framing this as assistance rather than automation is the right call - but it also means the real test will be in how the handoffs between AI recommendations and human decisions actually work in practice.

By the way, if you're building a project or startup around tools like this and need a fast way to get a web presence up, check out 60sec.site - an AI website builder that gets you from idea to live site in under a minute. Worth bookmarking.

Read the full story →

$875M is a serious commitment, and the framing of AI as a decision-support tool rather than a full replacement is the right instinct - but the FAA's track record on large technology projects means this one is worth watching closely as it rolls out.

Data center company Crusoe has raised $3.9 billion in a new funding round, valuing it at $30.9 billion. The company is using the capital to build both massive conventional data centers and smaller, modular 'AI factories' - essentially purpose-built compute units that can be deployed more flexibly than traditional hyperscale facilities.

The modular angle is interesting because it points to a real constraint in the AI infrastructure build-out: grid capacity. Fitting a 100-megawatt data center into an existing power grid is genuinely hard, and smaller deployable units let companies get compute online faster in locations that couldn't support a full hyperscale build. We've been tracking how the AI infrastructure race is accelerating - this is another data point in that trend.

Read the full story →

A $30.9B valuation for a data center company is a sign of how much the market believes AI compute demand is not slowing down - even as the AI slowdown debate dominates the headlines this week.

A new report covered by The Verge warns that e-waste from AI data centers has been vastly underestimated. By 2050, it could amount to enough discarded hardware to fill 23 million 40-foot shipping containers - roughly enough to circle the entire planet six times if lined up in a row.

The reason previous estimates were so much lower: they didn't account for all the supporting infrastructure that gets replaced alongside the GPUs. Cooling systems, networking equipment, power supplies - all of it turns over faster than standard enterprise hardware because AI workloads push components far harder. The environmental impact of the AI build-out is a story that keeps getting bigger.

Read the full story →

The AI industry is spending billions on compute but almost nothing on what happens to that hardware when it's obsolete - and 23 million shipping containers worth of electronic waste is a number that is going to start showing up in policy conversations very soon.

The answer is about 45,000 flights! The FAA handles roughly 45,000 flights on an average day across US airspace, which is why a decision-support AI that can process complex routing scenarios faster than a human controller is genuinely valuable - and why errors in that system have stakes well beyond a wrong autocomplete suggestion.

Today's Microsoft story got me thinking: have you ever changed how you use an AI product once you found out how its training data was collected? Hit reply and let me know - I read every single response and it genuinely shapes what we cover next.

That's it for today - a genuinely weird mix of a story. Napster cloning teachers, Microsoft calling its own industry a theft operation, the FAA going all-in on AI, and a data center e-waste number that should probably be in more conversations than it is. See you Monday with more. And if you want to browse everything we've covered, dailyinference.com has the full archive.

Know someone who wants to keep up with AI in 5 minutes a day? Share your link below. 1 referral unlocks the AI Tools Starter Kit, and 5 referrals unlock the AI Insider Briefing.