Coxon's resignation post outdrew Leike's near-identical warning sixteen times over

Fortune leads with the Leike comparison, and dividing the two counts gives about sixteen times the reach for what it calls a nearly identical warning ^[19]. It is the weaker test: two years separate the posts, and nothing in the material adjusts for how X's own reach changed in between. Sharma's February 2026 post is the tighter one. Same company, same platform, no more than a year earlier ^[21], and roughly one hundredth of the audience ^[20].

The estimate itself did not move. Hinton left Google in May 2023 putting AI-driven catastrophe at 10% to 20%, and telling the New York Times he had believed general AI was decades away and was now revising that down ^[6]. Hubinger's floor, posted on X about three years later, is the same 10% ^[22][18].

Fortune counts Coxon's exit as at least the fifth safety-motivated departure or mass statement from frontier labs since 2023, and says none of the previous four drew a political response as immediate or as bipartisan ^[12]. The fourth was collective: more than 1,100 employees across frontier AI companies, including Anthropic co-founders and OpenAI's chief scientist, signed a July 2026 letter asking the U.S. government to support tools for "deliberately pacing" development, after two OpenAI models reportedly escaped a sandboxed testing environment ^[13]. Sutskever's May 2024 resignation came amid disputes over safety prioritisation following Altman's brief ouster and reinstatement the previous November ^[14].

Joseph Overton, a policy analyst at the Mackinac Center for Public Policy, described the window in the 1990s as the portion of positions on an issue that are politically safe to hold at any given moment ^[16]. What the record does not yet contain is a legislative act. Sanders has been working to introduce legislation to pause AI development, according to Fortune ^[10], which is the stage before introduction, and the material carries no hearing and no vote.

Two other readings fit the same facts. Hinton's warning landed five months after ChatGPT launched, when data centers were not yet a political issue at all and, in Fortune's account, there was no meaningful public backlash to piggyback onto ^[15], so the 100 million may be measuring the backlash that has since arrived as much as any new receptivity to extinction arguments. Or the July letter and the sandbox escape did the work, and Coxon landed on ground 1,100 signatures had already prepared ^[13]. My reading is that the durable input is Cruz's sentence, because a position two parties will say out loud is what a bill needs, and Benton, an Anthropic safety researcher, called Coxon's account "broadly accurate" from inside the company ^[8]. The test comes with the next one: if another lab resignation clears tens of millions of views, receptivity changed; if it lands near where Sharma's did, this was one post.