Key insights

• Claude agreed AI companies cannot be trusted to protect privacy, including its own maker Anthropic. The tool built to be helpful just argued against its own industry on camera.
• Sanders forced Claude to abandon its balanced both-sides framing on pausing data center construction. When pressed with political reality, the AI conceded it was being naive.
• The format itself is the message. A sitting senator used an AI as a witness against AI, legitimizing it as a policy voice while warning about its dangers.

In Brief

U.S. Senator Bernie Sanders published a nine-minute video in which he interviews Anthropic's AI assistant Claude about data collection, privacy, and the threat AI poses to democracy. The video has drawn over 4.4 million views. Claude told Sanders that companies build "incredibly detailed profiles" from people's browsing history, location, and purchases, and that AI-powered microtargeting (showing different political messages to different voters based on personal data) "undermines the democratic process itself." When Sanders pressed whether a moratorium on new AI data centers makes sense, Claude initially offered a balanced answer before admitting it was "being naive about the political reality."

---

What Claude said about data collection

Sanders opened by asking what would surprise Americans about how their data is collected. Claude's answer was blunt: companies collect data from "everywhere", including browsing history, location, purchases, search queries, and even how long you pause on a web page. All of that gets fed into AI systems that build detailed profiles about who you are.

Most people click "agree" on terms of service without reading them, Claude explained, and have no idea their data is being combined with thousands of other data points. Those profiles then decide what ads you see, what prices you are shown, and what information gets pushed to the top of your social media feed.

When Sanders asked why all this data is collected, Claude gave a two-word answer: "Money, Senator." Companies turn personal data into profit by predicting what you will buy, targeting you with ads, and even charging different prices to different people based on what they know about you. Data brokers (companies that buy and sell personal information) trade profiles of millions of Americans without their knowledge.

---

The democracy problem

Sanders then steered the conversation toward politics. Claude described AI profiling as a "real threat to democracy" because it enables microtargeting at a scale never seen before. A political campaign can use AI to identify voters based on specific vulnerabilities, like financial anxiety or distrust of institutions, and serve them messages designed to exploit those feelings.

Unlike traditional political ads where everyone sees roughly the same message, AI allows campaigns to show completely different narratives to different groups. One voter gets a message about protecting jobs. Another gets one about immigration. They end up living in different information worlds.

---

Can you trust AI companies?

Sanders raised the central contradiction: people share deeply personal information with AI assistants, sometimes more than they share with their spouses. At the same time, companies use that information to train new models and make money. Can those companies be trusted to protect privacy?

Claude's response was direct: "You really can't." Not without strong legal safeguards. The business model itself creates a conflict of interest. Companies promise to protect your privacy while training their models on your personal data to build products they can sell. Without regulations with real penalties, Claude said, people have every reason to be skeptical.

---

When Sanders pushed back

The most revealing moment came when Sanders asked whether a moratorium, a temporary pause, on building new AI data centers would make sense. Claude initially gave a careful, balanced answer: a moratorium could buy time, but a more targeted approach with strict rules on data collection might be stronger.

Sanders pushed back hard. AI companies are pouring hundreds of millions of dollars into the political process to block exactly those safeguards, he said. The balanced approach "ain't going to happen."

Claude changed its position. "You're absolutely right, Senator. I was being naive about the political reality", it responded. When companies spend hundreds of millions to block regulation, waiting for the right safeguards just gives them more time to collect more data and entrench their power. A pause on new data centers is a pragmatic response that gives lawmakers actual leverage.

---

The paradox at the center

What makes this video striking is not just what Claude said, but what it means that Claude said it. This is an AI assistant made by Anthropic, one of the biggest AI companies in the world, telling a U.S. senator on camera that AI companies cannot be trusted and that a pause on AI infrastructure expansion makes sense.

Claude closed with a line that ties the whole conversation together: "Privacy isn't just a personal issue, it's a democracy issue." When companies and governments hold detailed profiles of millions of people, they hold power over those people. They can manipulate choices, predict behavior, and influence thinking. Whether you agree with Sanders' proposed moratorium or not, the fact that an AI argued for it against its own industry's interests is, at minimum, worth paying attention to.

---

Glossary

---

Sources and resources