🚨 Google's AI Is LYING About People

Google has long argued its AI just surfaces what's already on the web. A German court just rejected that, holding the company responsible for false scam claims its AI Overviews invented about two publishers, in what's believed to be the first ruling of its kind. The disclaimer shield is starting to crack. Let’s dive in!

In today’s insights:

LATEST DEVELOPMENTS

Evolving AI: A German court just held Google responsible for scam claims its AI Overviews fabricated.

Google argued its AI only surfaces what's already on the web, and a German court wasn't convinced. The publishers had flagged the false scam claims and been brushed off. With the complaint unresolved, they went to court, and the judges ruled that an AI summary is original content Google authors and stands behind, not a neutral set of links, which strips away the legal cover ordinary search results enjoy. A temporary injunction now blocks the false claims, in what's believed to be the first time a court has held an AI company responsible for its AI's defamation. Google has said it is studying the decision and weighing its response.

Google puts that AI summary right at the top of your search, and most people read it and move on without ever clicking through to check. When it gets your name or your business wrong, you've had nowhere to turn, since the company could shrug and point at a disclaimer. A court has now put the responsibility back on the company that served the answer. The next bad summary about someone finally has a place to go.

Most AI tools ask you to trade your data for intelligence. Norton Neo doesn't. It's the first safe AI-native browser built by Norton, and it gives you powerful built-in AI without handing your privacy over to get it. Search, summarize, and write with AI built directly into your browser. Your data stays yours. Your context stays private.

Built-in VPN, anti-fingerprinting, and ad blocking come standard. No add-ons. No setup. No compromises.

Fast. Safe. Intelligent. That's Neo.

Download Norton Neo

Evolving AI: Anthropic's CEO wants binding rules that could block its own future models.

Amodei lays it out in a new essay called Policy on the AI Exponential. For three years Anthropic pushed disclosure and transparency. Now he says the risks got real enough to need teeth. He points to Claude Mythos Preview as the proof, the model that scrambled the cyber landscape and showed what frontier AI can do. So he wants outside auditors and a government able to block unsafe releases, his own company's included. The plan goes well past the light-touch White House order from early June.

The reason that matters to anyone outside the lab is that the same tech writing code and chasing new drugs can also probe weak points in banks, infrastructure, and national security, and the danger is here now, not theory. Amodei's pitch is that nothing powerful reaches the public until it clears a safety check first, the way a plane gets cleared before you board. Slower releases, fewer surprises landing on you.

AI changed what senior engineering means. Gauntlet is built for engineers who want to lead that change — not watch it. The most effective way to become AI-first. No tuition. No tutorials. Just the work. Apply now.

Cohort is filling up now

_{Must be a US citizen to qualify.}

Evolving AI: Google's new open model DiffusionGemma writes text up to 4x faster.

DiffusionGemma delivers up to 4x faster generation on a single GPU, and that speed shows up in the raw throughput numbers. It runs past 1000 tokens a second on an H100 and 700 on an RTX 5090, all on hardware you can actually buy. It gets there by refining the text over several quick passes, sharpening a rough draft the way an image generator clears up a blurry picture. The result is fast, interactive local work like in-line editing and code.

Try it now on Hugging Face

If you have ever watched a local AI model crawl out one word at a time, this is the fix. A free model this fast and small means real-time editing and coding tools on your own machine, no monthly fee or cloud account attached. For developers who tinker at home, a fast model that costs nothing is a rare find.

Claude Fable 5 just played Pokémon FireRed from start to finish using nothing but raw game screenshots, no maps, no navigation aids, no harness translating the game state into text. That harness was the whole story with earlier models, so watching a vision-only run clear the game is a clean marker of how far general visual reasoning has come in a year. It is a timelapse, but it is worth a minute to see an agent navigate an entire game world the same way a person glancing at the screen would.

QUICK HITS

📈 Oracle's Cloud Infra Grows 93%,Free Cash Flow Goes Negative $23.7B.

💼 Amodei Floats Universal Basic Income as Labor Disruption Looms.

🛡️ Anthropic Floats Free Mythos Credits to All 50 US States.

🧬 Cathie Wood Backs Cellares' Robotics-and-Software Cell Therapy Platform.

🏥 Viking Global Leads $100M Series E for GT Medical's GammaTile Implant.

 *partner link