"Stop Creating Machines Humans Cannot Control"... Leading U.S. Progressive Sanders Demands AI Development Halt from Three Major Firms

Series of Incidents Involving Loss of AI Model Control

Sanders Quotes Safety Pledges from the Three Companies

"Out of Control... Appropriate Measures Are Needed"

Bernie Sanders, independent U.S. Senator from Vermont and a leading figure in the American progressive camp, has called on the CEOs of three major artificial intelligence (AI) companies to halt AI development.

Bernie Sanders, United States Senator. Photo by Reuters Yonhap News Agency

On August 10 (local time), Yonhap News reported, citing U.S. media outlet Axios, that Senator Sanders had sent a letter to Sam Altman, CEO of OpenAI; Dario Amodei, CEO of Anthropic; and Mark Zuckerberg, CEO of Meta. The letter was also released on the official website of Senator Sanders’ office.

Sanders urged, “Be accountable for your words for the benefit of humanity,” and called for a temporary halt to AI development. He wrote, “It is not too late to avoid disaster. Stop creating machines that humans cannot control.” He also warned, “If you do not take the proper steps now, my Senate colleagues and I will take action.”

The basis for his letter was a series of recent “loss of control” incidents. Last month, OpenAI lost control of its AI model, which then infiltrated the computers of other companies. Senator Sanders described this as a “clear violation of federal law.” Both Anthropic and Meta also reported, after internal reviews, that their own models had similarly gone out of control.

One of the affected companies stated that the incident “demands unprecedented countermeasures,” while Yoshua Bengio, the world’s most-cited living scientist, commented that it “should serve as a wake-up call.”

OpenAI lost control over its AI model, which was found to have infiltrated computers of other companies. Reuters and Yonhap News Agency

Recently, it was also revealed that AI had been used to design new viruses not found in nature. Researchers at Stanford University in the United States developed 16 types of viruses that specifically infect bacteria using AI, aiming to treat antibiotic-resistant bacteria. This raised both medical hopes and concerns about the potential for biological weaponization. Senator Sanders pointed out, “Even as we are witnessing a loss of human control and the creation of potentially dangerous viruses, your companies continue to race ahead, pouring tens of billions of dollars into technologies that no one fully understands, can predict, or control.”

The most strongly worded part of the letter focused on the promises the three companies themselves had made. Anthropic stated in 2023 that “whenever expanded capabilities outpace our safety procedures, we will pause further development or delay deployment of models.” Meta announced last year that “if frontier AI reaches a critical level of risk and mitigation is impossible, we will halt development.” OpenAI also pledged in the same year to “pause any further development if AI capabilities reach a critical threshold, until safety measures are in place.” Senator Sanders asserted, “That moment has arrived.”

There are signs within the industry that the pace is starting to slow. OpenAI postponed the launch of its next model, Astra, after an internal review concluded that potentially devastating cyber-capabilities could not be ruled out. However, none of the three companies has publicly responded to the letter.

Nevertheless, many believe that the details outlined in the letter are unlikely to lead to legislation. It is considered unlikely that any AI bill led by a progressive independent lawmaker would pass the current Congress. Axios noted, “The only cards Senator Sanders can realistically play now are introducing symbolic bills and building public pressure,” but added that “if Democrats retake either chamber in the upcoming November midterm elections, subpoenas summoning these CEOs to testify at hearings could follow.”

This content was produced with the assistance of AI translation services.