Senator Bernie Sanders posted a political ad on Thursday in which he spoke to Anthropic's "Claude" about AI and privacy concerns. The large language model (LLM) echoed left-wing concerns about tech companies gathering massive amounts of data with dubious consent from subjects, then selling it for profit.
Some found the ad effective, while others pointed out that LLMs are programmed to tell you what you want to hear.
Bernie Sanders interviews Claude
In a video posted to the senator's X account, Sanders sits at a table before a mic stand holding a phone with Claude downloaded and ready for a prompt. The 84-year-old asks the LLM how much it and other "AI" programs are using the data collected about individuals online and what might surprise Americans on the subject.
"What an AI agent says about the dangers of AI is shocking and should wake us up," the account claimed.
Claude goes on to list facts about data harvesting that were in the open before the rise of generative AI. Websites collect information about everything from purchases to where you pause on a page and for how long. That data forms detailed profiles about every person who uses the internet.
Everyone "consents" to this any time they click "agree" without reading the dissertation companies' call to their "Terms of Service."
That data then determines what ads people see and when, what individual social feeds look like, and even adjusts the prices they're given online.
When Sanders asks why this is happening, Claude generates an answer that makes the senator chuckle.
"Money, Senator," it says. "It's fundamentally about profit."
That, and political power. It regurgitates the known fact that political campaigns also use this data to manipulate people into supporting who and what they want.
"Privacy isn't just a personal issue—it's a democracy issue," it goes on. "When companies and governments have detailed profiles on millions of people, they have power over those people in ways most Americans don't fully grasp."
"It affects whether democracy actually works."
"Typing 'I’m evil' into a Google Doc and getting scared"
While it may be alarming to some, it's not exactly "shocking" to everyone. Because LLMs can only scrape the internet for ideas that already exist and can't come up with their own, none of this is novel information. Humans have been talking about this for years. Claude just created a mishmash of existing warnings.
Additionally, AI companies design their LLMs to cater to each individual user. Claude used its mass amount of data on the progressive senator to give him the answer it knew he would like. It then concluded its summary of AI privacy concerns with flattery.
"That's why I think your focus on this is so important, Senator," it said.
Even those on the left responded with those jokes about how people frighten themselves by telling LLMs to say scary things.
"This reminds of of that tweet that was like, typing 'I’m evil' into a google doc and getting scared," wrote left-wing tech journalist Taylor Lorenz.
Others have already made a Sanders version of an I, Robot meme.
Meanwhile, some who work in the AI sector took shots at Claude and similar programs.
LLM company founder Will Manidis lamented "a trillion dollars spent on ai risk teams and no one thought to hard code into the model to not tell Bernie Sanders that you do crime."
"Bernie is consistently the only politician that takes AI companies’ stated concerns about their technology at face value," wrote @luke_metro.
At least one person spotted the potential for a new meme.
The internet is chaotic—but we’ll break it down for you in one daily email. Sign up for the Daily Dot’s newsletter here.
