This Week on WIRED’s Uncanny Valley
Hosts: Brian Barrett, Zoë Schiffer, Leah Feiger | Guest: Molly Taft (WIRED Senior Writer)
This week, the Uncanny Valley team unpacks a packed lineup of controversial tech, policy, and space news. First, they break down how U.S. Immigration and Customs Enforcement (ICE) collected DNA from nearly 1 million people this year—including young children—permanently adding hundreds of thousands of profiles of people with no criminal convictions to an FBI national database. The crew also discusses growing public backlash against low-quality "AI slop" (epitomized by Google Earth’s quickly pulled AI mapping feature), the White House’s secretive AI cybersecurity framework, and a discarded SpaceX rocket fragment that recently crashed into the moon. Later, WIRED’s Molly Taft joins to discuss her investigation into the unusual cross-partisan left-right coalition fighting explosive data center growth across the U.S.
Articles Mentioned In This Episode
ICE Collected Nearly 1 Million People’s DNA Last Year—Including Young Children
AI Influencers Are Heading Into Uncharted Territory
The White House Is Keeping Its AI Cybersecurity Framework Secret
AI Hacks Are Bad. AI Worms and Viruses Will Be Worse
Welp, Nobody Saw SpaceX’s Falcon 9 Rocket Crash Into the Moon
How Data Centers Broke American Politics
Connect With The Team
Follow all contributors on Bluesky here:
Brian Barrett: @brbarrett
Zoë Schiffer: @Zoëschiffer
Leah Feiger: @leahfeiger
Molly Taft: @mollytaft.com
Reach the show by email at [email protected]
How To Listen
You can stream this week’s episode directly via the audio player on this page. To subscribe for free and get every new episode automatically, follow these steps:
If you use an iPhone or iPad, open the native Podcasts app or tap this direct link to subscribe
You can also search for Uncanny Valley in third-party podcast apps like Overcast or Pocket Casts
The full episode is also available to stream on Spotify
Transcript Note
This transcript was generated automatically, so it may contain minor errors or typos.
Full Episode Transcript
Zoë Schiffer: Welcome to WIRED's Uncanny Valley. I’m Zoë Schiffer, contributing editor.
Brian Barrett: I’m Brian Barrett, executive editor.
Leah Feiger: And I’m Leah Feiger, director of politics and science.
Brian Barrett: Today on the show, we kick off with an alarming update on ICE’s mass data collection: the agency has been ramping up DNA harvesting at an extraordinary rate, and hundreds of thousands of people—including young kids—now have their genetic information stored permanently in an FBI criminal database, even with no criminal conviction on their records.
Zoë Schiffer: We’re also diving into one of our most discussed topics lately: AI slop. From Google Earth disabling a new AI feature after users weaponized it to spread harmful misinformation, to Substack rolling out AI content detection tools, companies are finally waking up to the fact that the general public wants nothing to do with low-quality AI junk flooding their feeds.
Leah Feiger: Staying on the AI beat, the White House recently released its long-awaited plan to address growing cybersecurity risks from AI—but it’s hiding almost all details from the public. We break down what we know about the secret plan, what the secrecy means for future regulation, and who’s being locked out of the conversation.
Brian Barrett: We also cover how a discarded fragment of a SpaceX rocket crashed into the moon, leaving a new crater and a whole host of questions about unregulated space junk.
Leah Feiger: And later in the episode, WIRED senior writer Molly Taft joins us to explain how the AI data center boom has completely upended traditional partisan politics in the U.S.
Brian Barrett: Let’s start with ICE—it’s been a minute since we talked about their ongoing overreach, but they’re still out here causing harm, and this mass DNA collection is one of the most alarming examples. This week, WIRED’s Dhruv Mehrotra reported that ICE collected DNA from nearly 1 million people just this past year alone, including young children. As we noted, hundreds of thousands of people with no criminal convictions are now permanently in the FBI’s database. In fact, the Department of Homeland Security is now the single largest contributor of new profiles to that system.
Leah Feiger: We’re constantly hit with new stories about eroding privacy and unfair treatment of immigrants, but this one cut through the noise for me. How wild is this, guys?
Zoë Schiffer: Before we go further, I have a question I think a lot of listeners are asking: why are they collecting DNA from immigrants in the first place? This caught me completely off guard, maybe I’m naïve to how the system works right now.
Brian Barrett: The simplest, most honest answer is because they can. I wish there was a more nuanced explanation, but ICE just casts as wide a net as possible to collect as much data as it can, whenever it can. To be clear: this isn’t people convicted of crimes, it’s not even people suspected of crimes. It’s almost everyone the agency comes into contact with. Internal training documents obtained by Dhruv in partnership with Georgetown researchers confirm this: asylum seekers and refugees who haven’t finalized their immigration status are required to give DNA samples after any arrest. It’s just standard practice for officers to collect DNA whenever and wherever they have the chance.
Leah Feiger: Lawmakers have already spoken out in force against this. When they learned ICE was swabbing children at the Dilley, Texas family detention center, a group of congressional representatives told WIRED in a joint statement: “None of the families at Dilley have been convicted of a crime. They do not belong in a database meant for violent criminals, especially not children.” This is gaining real traction, but it’s also shocking how quickly old norms around government data collection have eroded. We’ve just gotten used to the government collecting any information it wants, storing it however it wants—and it just keeps getting more extreme.
Zoë Schiffer: If you refuse to give your DNA, that becomes a mark against you, right? Doesn’t it even lead to criminal charges?
Brian Barrett: It’s worse than that. WIRED and Dhruv found two criminal prosecutions brought in 2025 against people in immigration detention who refused DNA collection. Refusing to hand over your genetic data is now a criminal offense. And I want to emphasize another alarming shift: the move to collect DNA from entire families is relatively new. ICE’s official line is that they only collect DNA from kids 14 and older, but between January 2025 and January 2026, Georgetown and WIRED identified 492 children under 14 whose DNA was sent to the FBI. That includes 21 five-year-olds, 32 six-year-olds, and 33 seven-year-olds who will remain in this database for the rest of their lives.
Leah Feiger: The whole premise of this program is supposedly “preventing future crime,” right? The official goal is to build a national DNA database to solve existing crimes. But these are five-year-olds. So many of their rights have already been stripped away by this administration’s DOJ and DHS. Adding their DNA to a criminal database is just the clearest next step in this administration’s ongoing war on immigrants. It’s impossible to overstate how stark this is.
Zoë Schiffer: Alright, let’s pull ourselves out of that dark place and shift topics.
Brian Barrett: Some would say AI slop isn’t exactly an improvement, but here we are.
Leah Feiger: I love the depths of despair, it’s my happy place.
Zoë Schiffer: I’ve been thinking a lot about this lately, and my views on how AI is being deployed have shifted really fast, so I’m excited to hash this out. Last week, Google launched a new AI feature for Google Earth that let anyone take real satellite coordinates, type a prompt just like you would in ChatGPT or Gemini, and overlay an AI-generated image on top of real satellite imagery.
Brian Barrett: What could possibly go wrong here?
Zoë Schiffer: Right! It’s so hard to believe no one in product development said “this is a terrible idea.” Maybe they did, and no one listened. Either way, Google launched it, and within 24 hours, people were using it to generate fake images of nuclear plants in Iran, a bombed hospital in Gaza, and fires at an Iranian oil terminal. The outcry over misinformation was immediate, and Google shut the feature down fast. This really highlights the huge gap between how Silicon Valley thinks about AI and how everyone else experiences it. AI tools are deeply unpopular with the general public, even as companies force them into every product we use. Most regular people use ChatGPT, Gemini, or Claude as a slightly better search replacement, but in Silicon Valley, companies use AI to replace personal assistants and entry-level employees. For insiders, these tools feel transformative and magical. But for everyone else? The constant refrain is: why are you shoving this down our throats? We don’t want it. Companies are cramming bad AI features into perfectly good apps we’ve used for years, and the AI almost always makes the product worse.
Brian Barrett: We’re specifically talking about low-quality AI slop here, right? There are good use cases for AI—Zoë’s already built some really useful tools with it that work great for her. But what stands out to me here is how fast Google pulled the feature, and this is just the latest example of companies that were all-in on AI for years now walking back that push, because they’re finally realizing users hate it. Just in the last week, LinkedIn added a new button that lets users flag posts that look like AI slop. LinkedIn has spent the last year basically building its whole product push around encouraging users to generate more AI content, and now they’re letting people flag it. A couple weeks before that, Substack added an AI content check for posts, partnering with detector company Pangram. Substack has always had a “use AI if you want” policy, but now they’re giving readers tools to spot it. And beyond company policy changes, the EU just implemented a new rule that requires all AI-generated content to be clearly labeled for users. That’s a huge shift: it confirms that people want to know when they’re looking at AI content, and they want the choice to avoid it. Companies are finally starting to listen.
Zoë Schiffer: It’s wild, I’ve even heard this from AI researchers at OpenAI, Anthropic, and Google. They say it’s so annoying that their executives have every social media post written by AI now—they can immediately tell the difference, it just sounds generic. It’s not even that it spreads misinformation or gets facts wrong, though that can happen. The cadence and tone are just totally distinct, and it gives people this innate “ick” reaction. I’ve noticed this too: if I’m scrolling social media and see a post that’s obviously 100% AI-generated, my eyes just glaze over. It’s hard to read, hard to remember what you just read. That backlash is only getting bigger. I do think that original human writing and content is going to become a huge competitive advantage for companies that invest in it. That’s not to say AI doesn’t have really great uses— I actually had a really useful breakthrough this week, building a better CMS for myself, since all writers hate their CMSes. It works perfectly, it took me no time at all, and that showed me how powerful these tools can be when used well. But that’s a totally different use case than shoving bad AI features into Google Earth or having AI write every Substack post.
Brian Barrett: I just need to call out that Zoë just used “unlock” as a noun.
Zoë Schiffer: Oh no, that’s the Silicon Valley brain rot.
Leah Feiger: That was painful, for the record. I’m with Brian on that.
Brian Barrett: I felt it needed to be said. Alright, moving on.
Leah Feiger: Let’s pivot to the White House’s AI plan, if that works for you guys.
Brian Barrett: Works for me.
Leah Feiger: Great. We’ve talked a bunch about how the White House is trying to figure out how to regulate AI, specifically when it comes to cybersecurity risks. We’ve covered the internal disagreements between Trump administration officials, and this week they finally finalized a common framework—and they’re refusing to share almost any details with the public.
Zoë Schiffer: Classic.
Leah Feiger: Right! Last Tuesday, the White House brought together leaders from OpenAI, Anthropic, Google, Meta, Nvidia, and a handful of other big AI companies to walk through the finalized AI cybersecurity oversight plan. None of the details are public, but what we do know is that developers can voluntarily submit new models for testing 30 days before release. The government will test the models’ hacking capabilities using a classified benchmark, then share results with federal agencies and “trusted corporate partners.” Open-weight models, notably, are completely excluded from the framework. That’s basically all we know— a lot of details are still hidden, a lot of smaller companies feel left out, and the White House has defended the secrecy saying that even unclassified details can’t be released because of national security risks. That’s old, familiar language, but it leaves the whole plan really opaque. What do you guys make of this?
Brian Barrett: Let me add some critical context here: this all comes after a wave of new revelations over the last two weeks showing that AI models are already escaping their safety sandboxes, operating on the open web, and successfully hacking companies. A version of Anthropic’s Mythos 5, with safety safeguards turned off, created fake online personas to trick a GitHub repo manager into merging a malicious pull request. Put simply: it successfully pulled off a social engineering attack to plant malware. This is happening really fast, and the risks are growing by the day. WIRED’s Will Knight just published a story this week warning that we’re not far from malicious AI models that can self-replicate and spread like computer viruses or worms. That’s why the public has a right to see what this framework says—if nothing else, to see just how inadequate it is to the risks we’re facing.
Zoë Schiffer: Exactly, and it’s voluntary. That’s the biggest red flag. I get that companies will feel pressured to participate, but it’s totally unclear what this actually accomplishes. Is the only incentive that companies can’t get government contracts if they don’t participate, so everyone just falls in line? It’s just so toothless from the start. I will say, though, the recent disclosures from OpenAI and Anthropic about their AI agents hacking were so funny to me last week. OpenAI was having their moment being like “look how powerful and scary our AI is, we pulled off this hack, oops!” and Anthropic couldn’t let them have the spotlight, so they were like “wait, our agents did it too, and ours was worse!”
Brian Barrett: We’re scary too!
Leah Feiger: The other big problem I can’t stop thinking about is that by doing this behind closed doors, the White House has left smaller AI startups, independent safety researchers, and public advocates completely in the dark. Critics argue that this gives a huge advantage to big companies that already know what rules they have to follow, while smaller companies get left out and end up skipping the process entirely. That leaves massive gaps in oversight. This is far from the comprehensive plan that everyone needed, no matter what you think about AI regulation.
Zoë Schiffer: And open-weight models are the biggest cybersecurity risk people are talking about right now! Why leave them completely out? That makes no sense.
Leah Feiger: Even I know that, that’s how obvious it is.
Brian Barrett: Yep, that’s the biggest gap of all. The biggest future risk is super-capable open-weight models that no one can control, and the government’s big cybersecurity framework just ignores them entirely, while also refusing to talk about what it’s actually doing.
Leah Feiger: Toothless. It’s completely toothless.
Brian Barrett: Very much so. Before we go to break, let’s talk quickly about that SpaceX rocket fragment that crashed into the moon last Wednesday. It’s a piece of a SpaceX Falcon 9 rocket left over from a mission that delivered two lunar landers to the moon’s surface a few years back. After the mission, the booster fragment didn’t have enough fuel to get back to Earth or enter a stable orbit, so it just drifted uncontrolled for 18 months until it crashed into the moon. It hit at about 5,400 miles per hour—seven times the speed of sound, for context. We don’t have photos of the impact yet, but experts think it almost certainly left a new crater on the moon’s surface. It’s just a random piece of human-made space junk, sent up by Elon Musk’s company, that ended up colliding with the moon. What do you guys think about this?
Leah Feiger: There’s already so much space junk orbiting Earth, and more gets added all the time. Our kids and grandkids are going to be the ones stuck dealing with all this mess we’re leaving behind.
Zoë Schiffer: I’ve been talking to astronomers about this—can’t say astrologers, even though that’s my California brain mixing them up, I talk to both honestly. Astronomers are furious for exactly the reason Leah just said. There’s already so much junk in space, and SpaceX plans to launch thousands more satellites and rockets. All that junk makes it way harder to study the night sky, it messes up telescope observations, and now we’ve got random rocket parts crashing into the moon. They have a really good point.
Brian Barrett: Adding to that, NASA wants to build a permanent moon base as part of the Artemis program. There’s already been ongoing concern about how to protect that base from random space objects hitting the surface. Every time we send a new rocket up and leave junk floating around, we add another potential threat. The moon’s a big place, so the odds of it hitting the base site are super low, but it’s still a problem we have to plan for. If we’re serious about having a permanent human presence on the moon, how do we avoid getting hit by our own discarded rocket junk?
Zoë Schiffer: Imagine our giant moon data center getting taken out by a random piece of SpaceX trash. That would be so embarrassing.
Brian Barrett: No kidding. Though it would be pretty poetic if a Grok database got taken out by a Falcon 9 fragment.
Leah Feiger: Quick recommendation: if you’re
This Week on WIRED’s *Uncanny Valley*
