They Helped Build AI, Now They're Protesting It

On Saturday afternoon, hundreds of protesters stood on the sidewalk outside Anthropic's headquarters in San Francisco before eventually marching to OpenAI and xAI,.

I heard about the event on Reddit and happened to be downtown checking out the tulips, so I decided to stop by and see what they had to say. As someone who uses AI regularly to do research and ask questions, I was curious to hear out the people who think what I'm doing is dangerous. I half-expected them to be Luddites, anti-technology protesters, or anti-capitalists with an axe to grind. I was wrong.

To my surprise, the organizers and speakers were AI researchers themselves. The march was led by Michaël Trazzi, a former AI safety researcher at Oxford's Future of Humanity Institute who led a three-week hunger strike outside Google DeepMind's London office last year. One of the speakers was Dr. David Krueger, a University of Montreal professor who trained under Turing Award winner Yoshua Bengio. Holly Elmore, a Harvard PhD in evolutionary biology and founder of PauseAI US, also spoke, as did a UC Berkeley statistics professor. Many of these people trained at the same institutions as the leaders now running the labs they were protesting.

The protestors don’t want a ban on AI. They also aren’t against current AI models or narrow uses of AI, such as recommendation systems or Google Translate. They want a conditional pause on frontier model development and the race to build increasingly powerful general-purpose systems. They want each CEO to publicly commit to halting frontier development if every other major lab does the same. In his essay, “The Adolescence of Technology”, Dario Amodei writes:

The march began at Anthropic because in February 2026, the company removed its own internal safety pause mechanism. This is also the same month it closed a $30 billion funding round at a $380 billion valuation.

Critics like Yann LeCun, who left Meta to found his own AI startup, call existential risk "complete B.S." and accuse safety-focused labs of using fear to build regulatory moats against open-source competitors. The AI CEOs benefit from both narratives: "our technology could change everything" attracts investors, while "it's so dangerous only we can be trusted" attracts regulators. The doom story and the hype story are two sides of the same coin.

One of my favorite findings of the day didn't come from the main stage at all, but from a flyer being passed around nearby. Distributed by the allied group StopAI, it warned that "Techno-fascism & AI threaten Jobs, Democracy, Dignity, Sanity, Creativity, Humanity, Reality, Freedom, Nature."

While the speakers on stage focused on extinction risk, the documented harms of current AI needs no speculation. There are now documented cases of chatbot-induced psychosis, suicides, and medical misdiagnosis, which was rated the number one health tech hazard of 2026. Voice cloning fraud is up 680 percent in a year. The majority of new web content is now AI-generated slop, and AI-generated code has caused one in five security breaches. None of this requires artificial general intelligence. Even if you don’t believe in the future risk of AI extinction, the harms already here affect everyone.

AI models are extremely useful for many tasks, but there are clearly many harms. Yet no CEO is going to voluntarily stop pushing the frontier. Powerful AI is already here. Who gets to decide how we live with it?