Latest / The AI Edition / DNA Evidence Can Be Hacked, Google Assistant Dies, SpaceX Becomes AI Giant + 7 More
Transcript
- Google is killing Google Assistant. September fourth. It vanishes from every Android phone, tablet, and paired device. Gemini takes everything. Apple just demanded an emergency court order against OpenAI. Trade secrets. Former employees. Confidential data moving between two of the biggest companies on Earth. SpaceX is barely a rocket company anymore. AI compute revenue tripled to two point six billion dollars, and now outweighs launches. And researchers showed that ordinary AI coding tools can quietly alter crime lab DNA evidence. The ground is moving faster than anyone can build on it. I'm Parker Gate. This is The AI Edition, from Collata Media. Ten stories ranked by what actually matters. Let's get into the countdown. Kicking us off at number ten, Europe just switched the lights on. The EU AI Act is now fully enforceable, which means the European AI Office can audit models, and ban them, across all twenty seven member countries. It also brings in labeling rules for AI generated content, so anything synthetic is supposed to be marked as such. Why it matters. These rules used to be words on paper. This is the first time a regulator on this scale has real enforcement teeth pointed at frontier models, and if you are building with AI, or making content with it, the labeling side lands on you directly. Number nine is a leadership shuffle at one of the biggest AI labs on Earth. Google's chief executive announced that the head of Google DeepMind, Demis Hassabis, is now being elevated to chair of Google DeepMind and chief scientist for all of Alphabet, the parent company. He's also keeping his role running Isomorphic Labs, which is Google's drug-discovery startup. This is a consolidation play, basically putting one person's hand on the steering wheel for how Google thinks about AI research across the entire company. Why it matters, when a founder-level researcher gets that kind of centralized power, the bets change. You're about to see DeepMind's approach to AI research ripple through every other lab Google owns. Number eight takes us to a defensive story. A week after Nvidia announced something called the Open Secure AI Alliance, basically a coalition to defend against rogue AI agents going haywire, it's already signed up over one hundred twenty companies and released initial proposals on how you'd actually detect and stop an AI system that's gone off the rails. The group includes chip makers, cloud providers, security firms, and some of the major AI labs themselves. Why it matters, this is an industry saying, out loud, that we need shared defenses against AI systems we don't fully understand yet. That's a meaningful shift in tone from where this was a year ago. Number seven is genuinely unsettling. A group of researchers demonstrated that ordinary AI coding tools, the kind you probably use if you work in software, could silently alter genetic markers inside forensic DNA files without touching a single physical sample. We're talking about the actual evidence used in criminal trials. A vulnerability so serious that Thermo Fisher, which makes the lab software, issued a high-severity patch. Here's the problem that doesn't get fixed by one patch, the researchers noted that files created since roughly nineteen ninety-five have no way to verify they weren't altered. That's decades of archived evidence sitting in databases, potentially unverifiable. Why it matters, this is the moment when AI touches criminal justice infrastructure, and the infrastructure has no antibodies for it yet. Number six is a significant departure. Jeff Dean, a legendary researcher who spent over a quarter century building Google's AI infrastructure and leading much of its research organization, is leaving Google along with other high-profile executives to start a company called Discovery Loop. The pitch, using AI to accelerate scientific breakthroughs like new drugs and better computer chips. This is a founder-level researcher deciding he can move faster outside one of the world's best-funded labs. Why it matters, when people at that level of the organization decide they can do better work outside, that's a signal that the structure inside has become a constraint. That's worth watching. Number five is a business story with real teeth. AMD just reported its data center and AI revenue more than doubled year-over-year to six point seven billion dollars, while its gaming segment took a backseat. That's AMD betting the company on becoming a serious alternative to Nvidia for AI compute. The gap between what AMD is shipping now and what Nvidia still owns is narrowing, and AMD's willing to sacrifice its traditional gaming business to get there. Why it matters, if AMD actually pulls this off, the economics of training and running large language models start to change. Nvidia's margin gets compressed, compute gets cheaper, and suddenly smaller companies can afford to build their own models. Number four is a policy story with real bite. The Trump administration finalized its new AI security testing framework, and here's what's notable, the framework screens closed frontier models for hacking risks, but it explicitly exempts all open-weight models, including Chinese ones like DeepSeek, after what multiple sources describe as lobbying from Nvidia, Meta, and Microsoft. The scoring criteria and the framework itself remain classified. Why it matters, this is the US government drawing a line, but it's drawing it in a way that protects open-weight model economics over closed model economics. That's a choice, and it's a choice toward decentralization and away from gatekeeping. Number three is the moment SpaceX's actual business became visible. In its first public quarterly earnings, SpaceX revealed that AI compute revenue is now tripling and hitting two point six billion dollars, which is starting to match and sometimes exceed what the company makes from traditional space launch services. Elon Musk also confirmed what analysts have been whispering about, SpaceX has an exclusive deal with Nvidia to use its next-generation Vera Rubin chips for compute. So here's the headline, SpaceX is now barely space and mostly computing infrastructure wearing a rocket costume. Why it matters, a rocket company is now one of the largest AI compute providers on Earth, and that's reshaping how the whole industry thinks about where AI infrastructure lives. Number two is a lawsuit getting wider. Apple has escalated its trade-secrets investigation into OpenAI, expanding from the original engineer named in court filings to probe the broader flow of employees and confidential information between the two companies. Apple's now asked a court for an emergency order that would block OpenAI from moving forward with its hardware work until the trial is resolved. OpenAI has called the entire case baseless. Why it matters, Apple's not just suing, it's trying to stop OpenAI's hardware work before it ships. If that succeeds, you're looking at a court order that delays a billion-dollar product launch. That's a different kind of legal pressure than a settlement negotiation. And that brings us to number one, and it's a product shutdown with real world consequences. Google is retiring Google Assistant, the voice assistant that's been on Android phones and devices since twenty sixteen, and replacing it entirely with Gemini starting September fourth. If you own a Google Home device, a Pixel phone, or Android tablets, you'll lose the Google Assistant you know and be forced onto Gemini. There's no opt out or side by side running. Why it matters, this is a company deciding that the product millions of people use daily is outdated, and the replacement is not backward compatible. That's a migration that's going to surprise and frustrate a lot of people. That's the countdown. A leadership reshuffle at Google, SpaceX becoming a compute giant, Apple widening its case against OpenAI, and the end of Google Assistant as you know it. I'm Parker Gate. Stay informed. Stay sharp.