A pair of organizations issuing warnings about the possible dangers of artificial intelligence are backed by allies of the technology firm Anthropic and its regulatory agenda.
The nonprofit groups Tarbell Center and Model Evaluation and Threat Research have emerged as key players in the public debate surrounding AI risks and potential policy, promoting AI-skeptical journalism and risk evaluations, respectively. The Tarbell Center has ties to Meta cofounder and longtime pro-Democrat donor Dustin Moskovitz and Skype cofounder Jaan Tallinn, while METR is connected to Moskovitz, Tallinn, and AI researcher Paul Christiano, a Washington Examiner review of public statements found. All three tech leaders played roles in Anthropic’s rise to success.
Anthropic CEO Dario Amodei raised eyebrows by promoting METR as an AI safety solution in an essay this month. He announced METR’s team would act as “third-party evaluators” assessing Anthropic’s tools and encouraging every major AI company to take this “key step.”
The Tarbell Center, for its part, runs a fellowship program placing tech reporters at legacy news outlets to publish coverage of AI, the Washington Examiner previously reported. The Washington Examiner itself has discussed placing a reporter in its newsroom with Tarbell in the past, interviewing prospective candidates, although nothing ultimately came to fruition.
“Tarbell exists to strengthen independent, rigorous journalism about AI,” a Tarbell spokesperson told the Washington Examiner. “Editorial independence is central to that mission. Tarbell does not determine what fellows cover, how they approach a story or what conclusions they reach and our funders have no editorial role in the journalism we support.
“The fellows supported by Tarbell produce a varied body of work,” the spokesperson said. “This includes reporting that reaches a wide range of findings and conclusions.”
The push comes as members of Congress fiercely debate how to ensure AI does not harm society, while Anthropic skeptics warn the company’s inner circle is vying for a monopoly over the public policy discussion to advocate its preferred restrictions on the technology.
Moskovitz and Tallinn were both early investors in Anthropic shortly after its 2021 founding, according to multiple reports. Christiano was an initial trustee for Anthropic’s Long-Term Benefit Trust in 2023 and founded Alignment Research Center, which has directly assisted Anthropic’s work, according to past statements from the tech researcher and Anthropic. Christiano later stepped down from Anthropic’s Long-Term Benefit Trust to join the Anthropic-backed AI Safety Institute, a regulatory advisory agency for the U.K. government.
Anthropic did not respond to the Washington Examiner’s requests for comment.
“We do not accept donations by frontier AI companies or their employees, and our funders have no say in the projects we work on,” METR President Chris Painter told the Washington Examiner. He pointed to a prior statement in which he said his organization’s work does not promote “doomer” narratives about AI.
“We share the concern that there won’t be meaningful independent oversight of these companies,” Painter told the Washington Examiner. “We don’t think any small group should have authority over what happens with AI, and our goal is to get information from inside these companies out into the public domain – external investigators can help bring more information to light (as we saw in our report on the Hugging Face [hacking] attack).”
METR’s website says nine AI developers have adopted its “Responsible Scaling Policies,” which help decide when continuing to develop certain systems would be “too dangerous” without protective measures. The group warns that third-party entities such as itself are needed because “it seems very plausible that advanced AI systems could pursue goals that are at odds with what humans want.”
Moskovitz donated in 2022 to Christiano’s Alignment Research Center, which was a precursor to METR — although the donation was restricted from benefiting METR directly, the nonprofit group told the New York Post. Tallinn, meanwhile, has given more than $150 million to hundreds of projects that include METR, according to Time magazine.
AI’s media watchdog
The Tarbell Center’s affiliations with Anthropic supporters similarly run deep.
The Tarbell Center’s largest donors include Coefficient Giving, a group led by Moskovitz, as well as Tallinn’s Survival and Flourishing Fund. The center also lists Anthropic cofounder and head of policy Jack Clark as a speaker in its fellowship program, promising “a variety of perspectives from experts in both journalism and AI.”
Nevertheless, the nonprofit organization‘s website says it aims to hold “the companies and people building it to account.” Critics have described Tarbell-backed journalists’ narratives about possible AI harms as exaggerated.
FOREIGN ANTI-POLICE TECH GROUP ANNOUNCES IT’S SHUTTING DOWN OVER TRUMP SANCTIONS
“As models grow more sophisticated, many experts fear they could aid terrorists in unleashing new pandemics, empower malicious actors to carry out devastating cyberattacks, or further enable authoritarian regimes to tighten their grip on society,” the center’s site reads. “More advanced systems could also upend the economy, potentially leading to mass unemployment or social unrest.”
The group emphasizes that it does not take direct donations from companies such as Anthropic or individual employees working for them.