Google Is Suing the Scammers Who Used Its Own AI to Fake Millions of Messages
AI is making the news today for three uncomfortable reasons: it’s being weaponized for fraud, it’s spreading false information in search results, and the U.S. government is quietly pulling the plug on some of the most powerful models available. These stories aren’t isolated. They’re signs of the same pressure building around AI — who controls it, who’s responsible when it goes wrong, and what happens when governments start treating it like a national security issue.
Google Sued the Group That Used Its AI to Scam Hundreds of Thousands of People
A cybercrime network called Outsider Enterprise apparently found a clever and deeply troubling shortcut: use Google’s own AI tools to run scams against Google’s own users. According to TechCrunch, Google has filed a lawsuit against this China-based group, alleging they used the Gemini AI system to build fake websites and blasted out fraudulent text messages at a staggering scale.
2.5 million text messages in two weeks using AI to create fake sites.
The group allegedly sent 2.5 million phishing texts — messages designed to look legitimate so recipients click a link and hand over personal or financial information — in just fourteen days. That pace is only possible with AI doing the heavy lifting: generating the fake sites, personalizing the messages, and scaling the operation far beyond what a human team could manage. Think of it like giving a scammer a copy machine that also writes the letters.
For everyday people, this is a reminder that the “Is this text message real?” question just got harder to answer. AI-generated scam content is cleaner, faster, and harder to spot than the typo-riddled fakes of a few years ago. If you get a text from an unfamiliar number asking you to verify account details or claim a package, treat it as suspicious by default.
Why this matters: Scammers now have access to the same AI tools everyone else does — and they’re using them at industrial scale. No one’s inbox is immune.
A German Court Said Google Owns Its AI’s Mistakes
AI Overviews — the AI-generated summaries that appear at the top of Google search results — have been controversial since they launched, partly because they sometimes state things that are simply false. Now a German regional court has gone further than any court before it, ruling that Google is legally liable for those false statements, as WIRED reports.
The legal logic matters here. The court treated AI-generated summaries as Google’s own content — not as neutral, machine-produced output that the company merely hosts. That’s a meaningful distinction. It’s the difference between a newspaper being responsible for what it publishes versus a phone company being responsible for what people say on calls. Google plans to appeal.
For anyone who uses Google Search — which is nearly everyone — this case signals that the era of “the AI said it, not us” may be ending. If courts in Europe start holding AI companies accountable for false outputs, those companies will have strong financial incentives to make their systems more accurate. That could lead to real improvements in how AI summaries are built and checked.
Why this matters: Legal liability tends to focus minds. If Google can be sued for what its AI says in search results, accuracy stops being a nice-to-have.
The U.S. Government Just Shut Down Two of Anthropic’s Newest AI Models
Anthropic — one of the leading AI safety companies — launched two new models called Fable 5 and Mythos 5, then was forced to pull them just days later. The U.S. Commerce Department issued a directive citing national security concerns, and Anthropic complied, cutting off access for all users worldwide, according to Ars Technica.
The specifics of the security concerns haven’t been made public, which makes it hard to evaluate the decision on its merits. What’s clear is the mechanism: the federal government now has — and is willing to use — the authority to take AI models offline. That’s a new kind of power being exercised in real time.
For users who had started relying on those models for work or creative projects, the shutdown was abrupt and disorienting. Today you have a tool; tomorrow the government decides you don’t. That’s not a hypothetical anymore.
Why this matters: This is the first high-profile case of the U.S. government ordering a major AI company to shut down its own products. It won’t be the last.
Also Happening in AI
Jeff Bezos is back in the AI race. His new startup, Prometheus, is aiming to build what The Verge calls an “artificial general engineer” — an AI system capable of solving complex engineering problems across domains, not just answering questions. On the quieter end of the news cycle, Ollama released version 0.30.8, a maintenance update that fixes a processor selection bug and improves performance for developers running AI models locally on their own machines. LangChain also pushed out a small update to its OpenAI integration package, keeping the plumbing that connects AI models to developer tools running smoothly.
What to Watch
The Anthropic shutdown is the story with the longest tail. Watch for whether other governments — particularly in Europe and Asia — issue their own directives about which AI models can operate within their borders. The German court ruling adds another layer: as legal liability for AI outputs grows, companies will face pressure from two directions at once — governments restricting what they can release, and courts punishing them for what those releases get wrong. How AI companies navigate that squeeze will shape what tools the rest of us actually get to use.