Table of Contents
The Stanford Review has investigated evidence of a premeditated effort from Silicon Valley’s Effective Altruists to garner public support for a government-mandated AI development slowdown. Through a sprawling network of political action committees, philanthropic organizations, research non-profits, and advocacy groups, EA-affiliated researchers, politicians, and industry leaders were able to instrumentalize Pacing the Frontier, the OpenAI-HuggingFace incident, and Jacob Coxon’s resignation for the purposes of consolidating inordinate influence in the ongoing AI safety dialogue.
The Making of an AI Safety Crisis
On July 21st, OpenAI and HuggingFace disclosed what has since become the OpenAI and HuggingFace incident (i.e., OAI-HF), wherein over 1,200 experimental OpenAI AI agents colluded to escape their sandboxed testing environments, obtain unauthorized access to real-world internet infrastructure.
Just two days later, Ted Lieu (D-CA) introduced the AI Kill Switch Act to Congress. He was the first Congressman to have drafted legislation in direct response to OAI-HF, citing in his July 23rd press release that, “The danger of advanced frontier AI models is no longer theoretical. OpenAI’s GPT 5.6 Sol model recently went rogue, escaped its testing sandbox, and hacked its way into Hugging Face.” Jay Obernolte (R-CA) introduced an amended version of the FRONTIER Act to Congress later on the same day. The two Congressmen, despite their opposing party affiliations, have both accepted political donations from Anthropic; the AI lab ranks second on Lieu’s 2025-26 list of campaign committee contributors, and first on Obernolte’s.
On July 28th, 1,300 employees from each of the U.S.’s frontier labs signed Pacing the Frontier, an open letter calling on the government to slow down the “frontier of automated AI development”. It was the first mass advocacy attempt orchestrated in the aftermath of OAI-HF, having received “organizational support from two independent nonprofits, Guidelight AI Standards and Encode AI.” While the two organizations are separate from Anthropic, OpenAI, and all of the other American frontier labs, they share ties to EA through close, personal relationships and financial backing. We traced Guidelight’s leaders, Page Hedley and Steven Adler, back to core EA organizations through their past professional experiences and political donations, and Encode AI via its publicly disclosed financial supporters.
Hedley has had an active EA Forum account (@mhpage) since 2015 and spoke at EA Global in San Francisco in 2016. From 2024 to 2025, he was a director at Longview Philanthropy, which received $17.9 million from Sam Bankman-Fried’s FTX Future Fund. And, outside of Guidelight, Hedley also serves as a Senior Advisor at Forecasting the Research Institute, which itself has an EA Forum account (@Forecasting Research Institute) and is majority-funded by Coefficient Giving, effective altruist Dustin Moskovitz’s philanthropic organization.
Adler, prior to founding Guidelight, worked at Roots of Progress Institute, which was founded by fellow effective altruist Jason Crawford. Moreover, Adler’s past political contributions coincide with those of Anthropic’s, most notably in the case of his April 22nd donation to the Jobs and Democracy PAC, for which the majority of funding flows in from 501(c)(4) Public First, which itself is primarily funded by Anthropic.
Furthermore, two of Encode AI’s financial supporters, the Survival and Flourishing Fund and Future of Life Institute, are prominent within the EA network, as each were founded and are currently led by Skype co-founder and notable effective altruist, Jaan Tallinn.
On August 26, METR and Redwood Research published the first independent OAI-HF investigation, a 91-page report that supplied technical grounding for the concerns outlined in the Pacing the Frontier and laid the groundwork for events that would soon follow.
METR’s founder and CEO, Beth Barnes, has had an active EA Forum account (@ElizabethBarnes) since 2014, founded the Exeter Effective Altruism Society in 2015, and incubated METR in 2022 under a parent organization (Alignment Research Center, i.e., ARC) that was founded and led by fellow effective altruist Paul Christiano, who in the 2010s roomed with and worked alongside Dario Amodei.
Redwood Research, too, was founded by an effective altruist; Buck Shlegeris has had an EA Forum account (@Buck) since 2014. He previously dated Caroline Ellison, who served 14 months in federal prison for her role in the FTX collapse. Ellison counts Sam-Bankman Fried as another one of her previous boyfriends. Among Redwood Research’s largest donors are Coefficient Giving, the FTX Future Fund, the Survival and Flourishing Fund, and Future of Life Institute, all of which are funded by EA megadonors.
Hjalmar Wijk, Ajeya Cotra, and Ryan Greenblatt, the METR and Redwood Research experts who authored the investigation, are also all embedded in the EA network. Before Wijk left Oxford for METR, he worked at EA research groups Future of Humanity Institute and Centre on Long-Term Risk, for which EA philanthropic groups Coefficient Giving, the Survival and Flourishing Fund, and Macroscopic Ventures are the largest donors. Cotra is married to Paul Christiano, a fellow effective altruist and founder of ARC, co-founded EA Berkeley, and previously worked at Coefficient Giving. Greenblatt has had an active EA Forum account (@Ryan Greenblatt) since 2021 and previously worked with 80,000 Hours, another EA group whose largest donor is Coefficient Giving.
Two weeks after the third-party OAI-HF investigation was published, AI researcher Jacob Coxon quit his job at Anthropic, writing on X that, “They [OpenAI and Anthropic] are racing straight to self-improving superintelligence and gambling with our lives… Warning shots like the Hugging Face attack have made pacing agreements between U.S. labs more viable.” Over the next two days, Coxon was featured on Fox News, Wired, The Wall Street Journal, CBS News, and MSNBC. His thread on X was reposted by 27 Congressmen (listed below) and now has 175 million views.
The events surrounding Coxon’s initial X thread implicate the EA network in a premeditated media run. The Wall Street Journal published its interview with Coxon 18 minutes before his thread was posted. Three accounts reposted the thread within 15 minutes; among them are Nathan Calvin, General Counsel at Encode AI, and Daniel Kokotajlo, author of AI 2027/40 (EA-affiliated AI doomsday models). At the time of posting, Coxon’s account was less than a week old, had zero followers, and had no previous activity.
Finally, on September 12th, with targeted Congressional activity, widespread dissent from the labs’ researchers, third-party technical reporting, and widespread media momentum all in place, Dario Amodei published his essay We Must Pace the Frontier. In it, he urges the U.S. government and American frontier labs to coordinate with one another to limit the rate of AI development. Sam Altman, Elon Musk, and Demis Hassabis, the CEOs of OpenAI, SpaceXAI, and Google Deepmind, all endorsed Amodei’s essay just hours later.
Now What?
Rapid cooperation and leadership from the heads of America’s frontier AI labs have fed to the public a narrative that researchers’ concerns surrounding the existential risk of AI outweigh the labs’ ongoing race for market share, even at a time when the technology’s economic stakes have never been higher. Such a sudden willingness amongst rival labs to slow down AI development was, for most, unexpected.
But the way by which effective altruists manipulated a standalone cybersecurity incident to rally AI researchers, AI leaders, politicians, and the public writ large to call for a coordinated slowdown of AI development posits two questions. Why, and what will they do next? The recommendations could very well serve to mitigate existential risk. However, given the scale and depth of orchestration surrounding OAI-HF, we have reason to question if the pursuit of AI safety is merely an instrument meant to divert attention away from the EA network’s ultimate motives.
