A majority of Americans would like AI development halted immediately. Now the companies behind the technology are reportedly preparing for the moment that frustration turns into something less polite.
Axios, which first reported the story, says executives at leading AI labs, among them leaders at OpenAI and Anthropic, are putting together standard operating procedures for dealing with a large-scale revolt aimed at their companies. The exercises are modeled on Pentagon-style “war games.” Teams step through hypothetical situations, such as major cyberattacks enabled by the labs’ own products, or a political uprising that follows a catastrophic event triggered by their models.
OpenAI’s shrug of a statement
Asked by Axios about the drills, OpenAI kept its answer low-key. The company said “OpenAI conducts preparedness exercises where teams discuss and work through a range of potential scenarios.”
“These scenarios are not treated as inevitable, but are meant to help us prepare for a variety of circumstances,” the spokesperson said.
Put plainly, the company is planning for a day when its products help cause a disaster serious enough to send the public into the streets, and it has described that in neat, corporate language. Neat language, however, reveals very little about how seriously people inside the company view the danger.
Anthropic stays quiet, but its record talks
Anthropic did not comment. That doesn’t mean the company is unconcerned about social unrest.
Reports in September said Anthropic was developing a surveillance apparatus to monitor anti-AI protest movements, an effort that included a “pre-crime” approach to gathering intelligence. A plan for handling a public revolt fits neatly into a pattern that already existed.
The numbers behind the nerves
Public sentiment goes a long way toward explaining why these labs are rehearsing such scenarios. In polling from late September, the Wall Street Journal found that 63 percent of US adults back an immediate pause on AI development. Separately, a Politico survey found that roughly two-thirds of Americans hold at least “moderate” anxieties about AI destroying humanity.
Figures like these don’t point to a fringe group. They point to most of the country.
When the scale of the backlash first became apparent, tech billionaires such as Mark Zuckerberg responded with tone-deaf media stunts meant to win over millions of skeptics. That strategy clearly failed, and contingency planning followed.
How worried should anyone be?
How real any of these threats are remains difficult to judge. No one outside the companies knows the scenarios in detail, or whether the procedures are anything more than a conference room exercise.
Still, consider who is behind the drills. They aren’t activists forecasting doom. The people getting ready for a public uprising against AI labs are the ones running those labs, and the fact that they’ve written “catastrophic event triggered by our models” on a whiteboard shows how clearly that 63 percent is being heard in the boardroom.