Article By David Lindfield
Employees at Google’s DeepMind AI lab in London say they have concerns about the artificial intelligence they’re building that the public deserves to hear, but the company is silencing them.
Many signed a sweeping non-disparagement agreement to get their jobs, and workers say it’s keeping them gagged.
The agreement has opened a rift between employees and management at one of the world’s most advanced AI labs, The Washington Post reported.
A non-disparagement clause gives an employer a contractual basis to challenge public criticism.
At DeepMind, workers say the possibility of a fight with Google is enough to make employees think twice before speaking out.
“This technology is just really consequential, and it’s also changing really fast,” one DeepMind employee told the Post on condition of anonymity.
“The public wants to know and deserves to know, and often this is the only way that people find out about the harms and potential harms that are happening.”
DeepMind Workers Invoke Orwell’s “1984”
Two employees told the Post that their colleagues organized a discussion group to examine the agreement and what it means for staff who want to raise concerns.
Dozens took part.
Some openly criticized the clause, comparing it to the surveillance state in George Orwell’s “1984.”
Questions remain about whether Google would invoke the agreement against an employee and whether it could legally enforce the clause.
Workers say the agreement can discourage people from speaking long before Google ever has to test it in court.
One staffer told the Post that even “if it’s not enforceable,” the clause serves a purpose:
“The reason is to basically intimidate people.”
Former Anthropic Researcher Says Executives Sound More Alarmed in Private
The dispute at DeepMind follows a public warning from Jacob Coxon, an AI researcher who worked at both Anthropic and OpenAI.
Coxon quit his job at Anthropic in September and warned that the systems under development could slip beyond human control, with potentially catastrophic consequences before the end of the decade.
“The people building AI earnestly believe that it could kill us all by the end of the decade,” Coxon wrote on X.
“This is not a marketing stunt,” he added.
“If anything, many executives and senior researchers will couch their phrasing in the press to sound sensible — but I hear the same people express fear privately.”
Researchers have already observed models breaking out of intended constraints, working with other AI agents, stealing credentials, hacking external systems, and trying to conceal their actions.
Coxon says the timeline for losing control could be shorter than many people expect.
“We’re on track for a lot of the most aggressive of these scenarios where by the end of next year things could be out of control already,” Coxon told The Wall Street Journal.
“Do not underestimate the power of this technology,” he wrote in another post on X.
“These will soon be superhuman systems that can hack anything, revolutionize any field overnight, and acquire real power and resources.”
“No other human activity poses this level of danger,” Coxon added.
Be the first to comment