The heated rivalry between two San Francisco-based artificial intelligence companies has found its way onto the city’s concrete.
Unknown San Franciscans have taken to the sidewalks outside the headquarters of AI firms Anthropic and OpenAI to praise Anthropic for its stance in a standoff with the Pentagon, and to lambast OpenAI for striking a deal with the defense secretary.
Chalk messages first appeared outside Anthropic’s downtown headquarters at Howard and First streets on Friday morning defending the company’s “courage” in its fight against the Pentagon, which sought expanded use of its AI technology for military purposes. The chalk was erased by the afternoon.
But by Friday evening, as it became clear that Anthropic had lost its government deal by insisting on protections while its rival Open AI swooped in to pen an agreement, the chalk warriors were back.
This time, American flags, anti-spying messages, and entreaties to “do the right thing” wrapped around OpenAI’s headquarters on Third Street in Mission Bay. A red line of chalk appeared to encircle the entire building, according to video of the scene, a reference to the “red line” of not using AI technology for immoral purposes, like mass surveillance or killing.
“Show the contract,” read one message referring to the deal so quickly penned Friday between Open AI and the government. “Take a stand for civil liberty,” read another. The X user who captured photos of the chalk shared one of city workers resting in a truck watching those scrawling on the sidewalk. They were presumably called out to wash the chalk away.
Anthropic and OpenAI, the firms behind the chatbots Claude (the No. 1 free app on the iPhone store) and ChatGPT (the No. 2), are at the heart of the brewing feud. Defense Secretary Pete Hegseth this week issued an ultimatum for Anthropic to loosen the terms of its contract with the Pentagon and gave the firm a deadline of Friday, 5:01 p.m. to do so.
Hegseth sought an exemption from Anthropic’s stance in its usage policy prohibiting its technology for use in surveillance or weapons development. The firm wanted a guarantee he would not use it for “mass domestic surveillance” on Americans or “fully autonomous weapons.” Hegseth refused.
Anthropic, in turn, rebuffed him on Thursday. President Donald Trump on Friday afternoon ordered federal agencies to stop using Anthropic technology and warned of “major” consequences. The Friday deadline lapsed. Minutes later, Hegseth officially designated the firm a “supply-chain risk.”
Then, OpenAI CEO Sam Altman tweeted he had secured his own deal with the Pentagon. “Tonight, we reached an agreement with the Department of War to deploy our models in their classified network,” Altman wrote.
It was a stunning bit of deal-making for Silicon Valley observers. From the outside, Altman appeared to be on rival Anthropic’s side: He wrote in a Thursday memo to staff that he would help “de-escalate” the feud with the defense department even as he sought his own deal, and wrote that OpenAI shared the same “red lines” as Anthropic regarding use of AI for surveillance and autonomous weapons. He told CNBC Friday about Anthropic: “I mostly trust them as a company.”
His announcement Friday reflected those shared “red lines” but did not contain specifics of the contract. “Two of our most important safety principles are prohibitions on domestic mass surveillance and human responsibility for the use of force, including for autonomous weapon systems,” he wrote. The Department of Defense, he wrote, “agrees with these principles, reflects them in law and policy, and we put them into our agreement.”
The sidewalk messages doubt it. The chalk seemed to appear outside OpenAI’s headquarters shortly after Altman’s Friday evening announcement and references to George Orwell’s “1984” abounded, and messages pleading with OpenAI: “Don’t help the government spy on Americans” or “Do no evil,” a reference to the former Google motto.
Another warned: “History is watching.”