Anthropic ex-researcher's AI warning trailed by old tweets

Jacob Coxon, who spent three years in pretraining at OpenAI and Anthropic before quitting in September, posted that AI "could kill us all by the end of the decade." Internet sleuths have since linked his current X handle to older archived accounts that spent 2021 through 2024 waving away climate change, Futurism reported.

Coxon's current account, @hilbertspaess, was created in 2026. But "the gumshoes found that there were two previous accounts on that platform under the same handle between 2021 and 2024," according to the report. Archived posts include "it's literally 2020 already. look at climate change headlines from the 1970s then look around. nothing is happening," and the plea "stopppp talking about 'climate change' pleasse." Another questioned whether a warmer planet might carry upsides: "why it all gotta be bad?" One line, "six million is a palpable taunt," Futurism describes as "an apparent incredulous allusion to the Jewish death toll in the Holocaust," which it calls "a hallmark of Holocaust denialism."

The resignation itself landed hard. Coxon's post, which argued frontier labs are "racing straight to self-improving super-intelligence and gambling with our lives," was viewed more than 70 million times and drew responses from Elon Musk, Dario Amodei and Sam Altman. Four researchers on our Who's Who radar amplified the Futurism piece. Almost immediately after it went viral, Coxon appeared on Fox News, where he downplayed climate concerns while emphasizing AI's rapid development as uniquely dangerous.

Whether Coxon has acknowledged authorship of the earlier accounts, and whether Anthropic has any comment on them, the report does not say.