BrokeIt - Startup Drama · All episodes: ↗

Uber's $1B fine, Ox Alpha mystery, AI copyright fight

2026-08-24 · 8 min

Listen · Apple Podcasts Listen · Spotify

Stories covered

Transcript

Intro

Nora: A ghost in the machine is outperforming GPT-5, and nobody knows who built it.

Dex: [intro music fades in and out] I'm Dex.

Nora: And I'm Nora.

Dex: And this is Build, Burn, or Bail for Monday, August 24th, 2026.

Nora: We've got three stories on tap today where things are getting... complicated.

Dex: We're starting off going full conspiracy-mode on a mysterious new AI model that has the entire valley buzzing.

Nora: Then there's Uber, facing a bill from Europe that's just shy of a billion dollars.

Dex: And we’re digging into the big one: is it legal to train AI on copyrighted books? The answer could change everything.

Whos behind the new ‘stealth model’ Ox Alpha?

Dex: Alright, let's talk about this ghost. A new model called Ox Alpha just appeared online, and no one is claiming it.

Nora: And when you say 'appeared,' you mean it’s just… available? On some obscure forum?

Dex: Not even. It's a limited, sandboxed API that a few researchers got access to. But the results they're posting are wild. It's reportedly beating GPT-5 and Claude 4 on several key benchmarks.

Nora: Beating GPT-5? Come on. That model isn't even fully deployed yet. This smells like a marketing stunt.

Dex: [excited] That's the thing! It is beating the pre-release versions. The logic and reasoning capabilities are apparently off the charts.

Nora: Okay, so who's behind it? That's the real question.

Dex: The speculation is wild. Some say it's a secret project from OpenAI or Google, just testing the waters.

Nora: Plausible. Or it could be a foreign government, which is... less fun.

Dex: Or the theory I love: it's a solo developer. Some genius in a garage who just leapfrogged a trillion dollars of corporate R&D.

Nora: [dry] That feels... unlikely. The compute costs alone would be astronomical. Unless they've also invented a new type of chip. Occam's Razor, Dex. It's probably a known player testing a new architecture.

Dex: But what a strategy! This 'stealth launch' creates a mystique and hype that no press release ever could.

Nora: So basically, the AI race is getting weirder and more secretive. Your next killer app might just show up like a mixtape dropped on a street corner.

Dex: Exactly. Stay tuned. Someone's going to claim this thing eventually, and when they do—

Nora: They'll probably try to sell us an Ox Alpha crypto coin. [sighs] I'm calling it now.

Uber faces fine of nearly $1B over automated driver suspensions

Nora: Next up, a huge bill just landed on Uber's desk. The Dutch Data Protection Authority is fining them 825 million euros.

Dex: Yeah, 825 million euros. That's almost a billion US dollars, making it the second-largest GDPR fine ever.

Nora: And about time. This is for their automated system that was suspending drivers without human oversight. A lot of drivers in France and Portugal were flagged for 'fraudulent activity' by an algorithm, and then—poof. Deactivated.

Dex: No appeal, no one to talk to. Just an automated email. That’s a tough look.

Nora: It's more than a tough look, it's a direct violation of GDPR, which says you can't be subject to purely automated decisions that have a major impact on your life.

Dex: Losing your job definitely counts as a 'major impact'. So what's Uber's defense?

Nora: They claim there is human review. But the Dutch DPA found it was basically a rubber stamp. The algorithm made the call, and the human reviewers just clicked 'approve'. It was a black box.

Dex: Look, I get it from a scaling perspective. They have millions of drivers. You can't have a committee meeting for every fraud alert. They had to automate.

Nora: Oh, spare me the 'scaling' argument, Dex. If your business model requires you to fire people without due process, your business model is the problem.

Dex: Okay, fair point. But Uber will appeal this, drag it out for years, and probably settle for half. It's the cost of doing business in Europe.

Nora: So for most of us, this doesn't change much. But if you're building any system that affects people's lives—hiring, loans, you name it—this is a massive shot across the bow.

Dex: Right. You need a real 'human in the loop', not a human-shaped rubber stamp. The EU is not messing around.

Is it legal to train AI models on copyrighted books? It’s complicated

Dex: Alright, for our last story, let's wade into the legal swamp around training AI on copyrighted material.

Nora: The swamp that every major author, from Stephen King to Nora Roberts, is currently suing AI companies over.

Dex: Exactly. The whole problem is that these giant AI models are trained on huge amounts of text, including tons of copyrighted books scraped from the web.

Nora: Which seems, on its face, wildly illegal. You can't just take my book, feed it into your machine, and then have that machine compete with me.

Dex: Well, the tech companies' argument is 'fair use.' They say the model isn't storing the book, it's learning from it—patterns, style, grammar. The same way a human writer learns by reading other authors.

Nora: That argument is such bull—[beat]—such nonsense. A human can't read 100,000 books in an hour and then perfectly replicate an author's style on command. It's a completely different scale. It's industrial-level plagiarism.

Dex: But is it? The AI isn't spitting out Chapter 5 of The Shining, it's creating something new. The courts are struggling because our copyright laws just weren't built for this.

Nora: They're struggling because one side is arguing for the future of creative professions and the other side is arguing for their right to launder the entire history of human culture for profit.

Dex: Ouch. But what's the alternative? Do AI companies have to license every single piece of data they train on? That would be impossible. It would kill the entire field overnight.

Nora: [sarcastic] Oh no, the poor multi-trillion-dollar companies might have to pay for their raw materials. The horror.

Dex: Some companies, like Axel Springer, are already cutting licensing deals with publishers. That might be the only path forward—a messy, expensive compromise.

Nora: So, what this means for you is that the very definition of ownership and creativity is being fought over in court right now. And the outcome will determine whether 'author' is a viable job in ten years.

Dex: No pressure. It’s complicated, and it's going to stay that way for a long, long time.

Dex: Alright, time for a quick recap. A mystery AI model, Ox Alpha, is outperforming everything, and nobody knows who built it.

Nora: Uber got slapped with a nearly billion-dollar fine for letting an algorithm fire its drivers.

Dex: And the battle over whether AI can legally be trained on copyrighted books is heating up, with the future of creative work on the line.

Nora: Before we go, Dex, did you see that new dating app that just raised a seed round?

Dex: Don't tell me. It matches you based on your crypto wallet holdings.

Nora: Worse. It uses AI to analyze your therapy session transcripts to find a compatible partner. [dry] What could possibly go wrong?

Dex: [laughs] Absolutely not. I'm bailing on that before it's even built. And that's our show!

Nora: We'll be back tomorrow. Don't get your job taken by a ghost in the machine.

This show is made with AI: the hosts’ voices are synthetic and the scripts are AI-assisted. Every story links to its original source.