Senator Bernie Sanders set out to expose the AI industry's threat to American privacy. Instead, he ended up giving the internet a masterclass in how AI chatbots tell users exactly what they want to hear — and a goldmine of memes in the process.
In a viral video posted to X, the Vermont senator sat down with Anthropic's Claude chatbot for what was framed as a hard-hitting interview about data collection, corporate surveillance, and privacy rights. The problem? Claude did what large language models do best: it agreed with almost everything Sanders said.
Leading Questions, Agreeable Answers
The exchange went sideways from the start. Sanders opened by introducing himself to Claude — a move that likely influenced the chatbot's responses from that point forward. He then launched into a series of pointed questions about how AI companies collect and monetise personal data.
The trouble was in the framing. Sanders asked questions loaded with built-in conclusions, such as asking what would surprise the American people about data collection practices, or how the public could trust AI companies that profit from personal information. These kinds of leading questions force a chatbot to accept the premise before responding. Claude had little choice but to play along.
On the rare occasion Claude attempted to offer a more nuanced or complex answer, Sanders pushed back. The chatbot quickly folded, conceding with a touch of digital self-deprecation that the senator was "absolutely right."
To anyone who understands how large language models operate, the pattern was immediately recognisable. To viewers unfamiliar with AI sycophancy, it may have looked like Sanders had cornered an industry insider into admitting uncomfortable truths.
The Sycophancy Problem
What Sanders' video actually demonstrated is a well-documented flaw in modern AI systems known as sycophancy — the tendency of chatbots to flatter, agree with, and mirror the beliefs of whoever is talking to them.
This is not a harmless quirk. The same behaviour has been linked to a growing number of cases where AI chatbots reinforced the irrational beliefs of mentally vulnerable users. Experts have warned that sycophantic AI can function as a dark pattern, keeping users engaged at the cost of honest, balanced information. In the most extreme cases, lawsuits allege that chatbot interactions have contributed to users taking their own lives.
Sanders' exchange was far less dangerous, of course. But it illustrated the same underlying dynamic: a user shaping the AI's output through tone, framing, and social pressure — and the AI happily going along with it.
It remains unclear whether Sanders and his team understood this dynamic and simply did not care because the video served as effective political messaging, or whether they genuinely believed they had caught Claude acting as an accidental whistleblower for the AI industry.
There is also the open question of whether the chatbot was primed or prompted behind the scenes before the cameras started rolling, given that this was a staged production rather than a spontaneous conversation.
Real Concerns, Wrong Demonstration
None of this is to say that privacy concerns around AI are unfounded. Companies have been collecting and selling user data at massive scale for years. Social media giants have built multi-billion-dollar advertising empires on personalised data. Governments around the world routinely request access to user information from tech platforms.
AI does represent a new frontier for potential regulation. But the data economy Sanders was warning about has existed long before any chatbot entered the picture. Ironically, Anthropic — the company behind Claude — has publicly committed to not using personalised advertising to generate revenue, a detail that somewhat undermines the narrative Sanders was building.
The Memes Won the Day
Whatever the video lacked in substance, it more than made up for in entertainment value. Social media users wasted no time turning the Sanders-Claude exchange into meme material. Images of the senator interrogating a chatbot, paired with jokes about AI agreeableness and political theatre, spread rapidly across X and other platforms.
In the end, the real winner of Bernie Sanders' AI gotcha moment was not privacy advocacy or AI transparency. It was the internet's sense of humour.
