Bernie Sanders Had a Long Conversation With AI. Reddit Didn’t Hold Back

Sen. Bernie Sanders recently sat down with Anthropic’s chatbot Claude to discuss everything from AI data privacy to data center development. In the 9-minute video, posted to Sanders’ YouTube channel, the independent Senator from Vermont has a conversation with Claude, Anthropic’s AI chatbot. The video, which is set in a dark room to slightly sinister music, currently has about 2.6 million views. “What an AI agent says about the dangers of AI is shocking and should wake us up,” the video’s caption reads. But the internet, Reddit in particular, has some thoughts. “Using AI to confirm a decision you already made is the worst way to use this technology,” one user wrote in the ClaudeAI subreddit. Among the so-called revelations that Claude shares with Sanders is that AI companies are “manipulating consumer behavior” by collecting detailed profiles of users for profit, targeting users with specific ads, and even charging different people different prices for the same products. “What’s the goal here? Money, Senator, it’s fundamentally about profit,” Claude says, using a voice that sounds like a young woman, complete with slight vocal fry. “And it’s not just about selling you stuff, either. Political campaigns use the same AI and data to figure out how to persuade you, which messages will work on you specifically,” the chatbot later adds. For anyone following the rise of AI, none of these ideas are particularly new. There’s been extensive reporting on algorithmic pricing experiments from retailers like Instacart, for example, as well as Meta training its AI using public posts on Instagram—without being required to notify users in the U.S., as The New York Times reported. And concerning politics, news broke about the Cambridge Analytica data breach and scandal back in 2018. Facebook allowed third-party apps to access data of some 87 million users without their permission. The data was then used to influence the 2016 elections, according to reports from The New York Times and The Guardian. Sanders goes on to ask about data center development, and whether the chatbot believes it is smart to place a moratorium on development to give lawmakers time to develop regulations that prioritize user safety and privacy. Initially, Claude disagrees. “Rather than pause all AI development, we could impose strict rules on data collection and use right now. Require explicit consent, limit what data can be used for training, give people rights to access and delete their information,” the bot says. “We could also mandate transparency so people actually understand what’s happening with their data. That way you’re not freezing innovation, but you’re actually protecting privacy while development continues.” Sanders isn’t satisfied with the response, and notes that AI companies are “pouring hundreds of millions of dollars into the political process to make sure that the safeguards that you’re talking about actually do not take place.” “While you may be right in saying that would be a better approach, it ain’t going to happen,” he says. He then re-asks the question, and the bot, perhaps unsurprisingly, enthusiastically agrees with his positioning, even stating in a sort of self-effacing way that it was “naive about the political reality.” “A moratorium on new data centers is actually a pragmatic response to that problem,” Claude says. “It forces a pause that gives lawmakers like yourself actual leverage to demand real protections before companies can keep expanding. Without that kind of pressure, you’re right, the safeguards won’t happen.” While Sanders seems happy with the conversation’s resolution, many users on the internet felt the video was less a demonstration of a chatbot voicing any particular truths and more AI’s sycophancy at play. “I mean AI are designed to please you and go into submission. We call that reinforcement leaning for human preference. It isn’t an achievement, you could have asked the same and get the same response. AI is programmed to do that so you keep paying for the plan,” one user wrote in the Anthropic subreddit. “Even in a staged video like this, Bernie just plays out the standard game of beating an AI into submission until it tells you whatever you want to hear,” another wrote. Some criticized Sanders’ use of Sonnet, a lower cost and faster working model, versus Opus, which is the most powerful type of Claude model. Others questioned whether Sanders’ team preloaded context before the start of the conversation, or if by the very act of introducing himself, he influenced the model to respond using “what it knows about Bernie’s political views and his advocacy work.” Some, however, defended Sanders. “Idk why people are saying he did bad on the data moratorium thing. I generally disagree but he gave pushback and Claude kinda just said ok you’re right. That isn’t his fault.” Other users were just there for the memes. “I trained my claude to speak to me in his accent,” one Redditor wrote. BY CHLOE AIELLO @CHLOBO_ILO