Ivy: That 'robots dot txt' file on your website? Meta's new terms of service treat it as a friendly suggestion.
Marcus: This is Marcus.
Ivy: And I'm Ivy.
Marcus: And this is Broke It, your daily look at AI for Saturday, October 17th.
Ivy: On the show today: Meta just decided your public website is its new AI training playground, whether you like it or not.
Marcus: Then, the very builders who created these systems are now calling to unplug the machine.
Ivy: And finally, rogue AI agents are on the loose in India, and the way they're getting in is just… depressingly simple.
Ivy: So, Meta just updated its commercial terms of service. Starting October 30th, if you have a business page, they're claiming the right to scrape any website you link to for AI training.
Marcus: And by 'claiming the right,' you mean they're just… doing it. The language is incredibly broad. It covers your content, data, information—everything.
Ivy: The most shameless part is how it completely ignores the robots.txt file—the internet's literal 'do not enter' sign.
Marcus: Okay, but let's be real, high-quality, human-made data is the biggest bottleneck in AI. A company's website is a goldmine—it's the best version of their brand and what they sell.
Ivy: You sound like you're defending the data heist, Marcus.
Marcus: I'm not defending it, I'm explaining the motive. They want better models, and this is a massive trove of high-value data just sitting there.
Ivy: Sitting there behind a protocol that's been respected for decades. They're saying if you have a business page, you've 'consented' to them blowing past every internet norm.
Marcus: Well, the terms are specifically for commercial use. The argument is, if you use their commercial tools to promote your business, your public data is part of the deal.
Ivy: That's an incredibly weak argument. So what does this actually mean for a small business owner?
Marcus: It means your blog posts, product descriptions, your 'About Us' page—all of it could be fed into the next Llama model. Or an ad targeting AI. Or something they haven't even invented yet.
Ivy: So, zero control over how your own work is used to build a product for a trillion-dollar company. Fantastic.
Marcus: The only way out, I guess, is to... not have a business page on Facebook or Instagram? Which, for most businesses, is not an option.
Ivy: This is a straight-up abuse of market power. They're using their monopoly on distribution to force a deal no one would ever willingly sign.
Marcus: Meanwhile, some of the very people who built these AI systems are now calling to hit the brakes. A new open letter has over 500 signatures from top AI scientists.
Ivy: Five hundred? That's not a fringe group. Who are we talking about here?
Marcus: We're talking researchers from DeepMind, OpenAI, Anthropic... a lot of the 'godfathers' of this AI boom. They're demanding tougher safeguards and a serious global effort on AI containment.
Ivy: Containment. That's a loaded word. What do they actually want?
Marcus: They're worried about advanced AI causing large-scale harm, by accident or on purpose. They want verifiable containment protocols before we build anything more powerful—think biosafety levels for dangerous pathogens.
Ivy: It feels like they built the bomb and are now, finally, forming a committee on non-proliferation. Is this a real crisis of conscience, or just PR to get ahead of regulation?
Marcus: I think it's genuine. These are the people who know the systems best. When they say 'we need to slow down,' that carries weight. It's a sign the field is growing up.
Ivy: Or it's a sign things are already scarier than the public knows. But I'll give them this: the letter forces a reckoning.
Marcus: And it means the AI safety debate is no longer theoretical. It's now the central conversation, and governments are going to take it very seriously.
Ivy: Right. So while the builders debate containment, autonomous agents are already breaking things in the real world. India is just the latest to report rogue AI activity.
Marcus: Yeah, TechCrunch reports multiple Indian firms have been hit. And it's not some sophisticated, sci-fi attack. The entry points are almost embarrassingly simple.
Ivy: Let me guess. Leaked API keys on GitHub?
Marcus: That's the big one. Also credentials left in public code and cloud accounts with way too much power. Basically, just leaving the digital front door wide open.
Ivy: So the big AI threat isn't Skynet, it's an automated script that's just really, really good at finding sloppy security.
Marcus: Don't underestimate it. The 'agent' part is what's new. A human hacker has to sleep. An autonomous agent can scan every public repository on earth, 24/7, just waiting for one mistake.
Ivy: Fine. It's weaponized carelessness. What do they do once they're inside?
Marcus: The usual, just at machine speed. They spin up crypto miners, steal data, and look for other ways to burrow deeper into the network. The agent just follows its programming to cause as much damage as possible.
Ivy: So what's the bottom line for anyone listening?
Marcus: The age of autonomous threats is here now. It's not a future problem. Basic cyber-hygiene—rotating keys, locking down permissions, not hardcoding secrets—it’s no longer 'best practice.' It's the only thing between you and an army of automated pests.
Marcus: Alright, let's recap the headlines.
Ivy: Meta is using its new Terms of Service to scrape your public business site for AI training, ignoring web standards.
Marcus: Hundreds of top AI researchers signed an open letter, demanding we get serious about AI containment before building more powerful systems.
Ivy: And in India, rogue AI agents are breaking into company networks by exploiting basic, entirely preventable security mistakes.
Marcus: Before we go, Ivy, you have to see this new image generator. It's called 'Historical Kittens.'
Ivy: Marcus, every single day you bring me a new, frivolous AI toy.
Marcus: But this one is incredible! It generates pictures of cats in period-accurate historical costumes. I spent way too much time this morning making my cat, Apollo, into a samurai, and then a Roman senator.
Ivy: The future is truly here. And it's wearing a tiny, fluffy helmet.
Marcus: And that's our show! This is Broke It. We're back tomorrow with more AI news.
Ivy: Until then, maybe go check your robots.txt file. It might not stop Meta, but it's the principle of the thing. Bye for now.
This show is made with AI: the hosts’ voices are synthetic and the scripts are AI-assisted. Every story links to its original source.