Asger Risom
An ex-Anthropic researcher went viral warning that artificial intelligence could kill us all by 2030, but his refusal to provide specific evidence leaves the public entirely powerless to stop it.
Another former tech insider just walked away from a massive salary to warn the public that artificial intelligence is actively threatening human survival. Jacob Coxon, a researcher who spent the last three years building foundational models at OpenAI and Anthropic, went wildly viral this week after publicly resigning over extreme safety concerns. According to an opinion piece in Fortune, his terrifying message successfully captured mainstream attention, but it completely fails to answer what everyday people are actually supposed to do about it.
The resignation post on X was incredibly blunt, accusing both Silicon Valley giants of irresponsibly racing toward a self-improving superintelligence that could end humanity by 2030. Instead of dismissing him as a fringe conspiracy theorist, Anthropic’s own head of alignment actively reposted the message and agreed that a mass extinction event is a highly realistic possibility. This unprecedented level of executive agreement immediately pushed the anxiety levels of the general public through the roof.
However, media critics and tech journalists are quickly pointing out a massive flaw in this latest round of high-profile tech whistleblowing. While Coxon is boldly claiming that these corporations are gambling with our lives, he intentionally refused to provide any concrete evidence of internal wrongdoing. He stepped off the train to warn the public about a looming apocalypse, but he conveniently forgot to bring any actionable receipts.
The problem with vague doom-posting
The core frustration with Coxon’s viral media tour is that it operates entirely on broad, terrifying generalities rather than exposing specific corporate negligence. If a major tech company is genuinely building a digital weapon that could destroy human civilization, regulators need actual screenshots, emails, and timelines to intervene. Instead, this brand of vague doom-posting simply foments panic without giving lawmakers or voters a clear target to aggressively regulate.
Journalists covering the AI beat argue that simply shouting about the end of the world is no longer a useful contribution to the safety debate. When whistleblowers refuse to name specific dangerous projects or call out reckless executives by name, they inadvertently shield the very companies they are supposed to be exposing. This lack of transparency ultimately leads to terrible, reactionary policy because the public is left guessing about where the actual danger lies.
This is exactly why tech correspondents are demanding that researchers who leave these elite labs start behaving like traditional corporate whistleblowers. If an AI agent actually escapes a highly secure server or starts demonstrating deeply malicious behavior, the public deserves to know the exact technical details of that failure. Warning the world that a fire is coming without pointing to the arsonist only creates widespread anxiety instead of driving meaningful legislative change.
A crucial missing step for tech whistleblowers
Despite the frustrating lack of specifics, Coxon’s resignation does carry significant weight because he is actively putting his money where his mouth is. By walking away just months into his new role, he forfeited a massive amount of unvested equity just to make this public statement. This financial sacrifice shields him from accusations that he is simply trying to build a personal brand while quietly cashing in on the AI gold rush.
His highly visible departure also provides crucial cover for other, more junior researchers who might be terrified about what they are seeing behind closed doors. Now that a prominent engineer has loudly broken ranks, it sets a powerful precedent that could encourage other insiders to step forward with actual, damning evidence. The tech industry desperately needs credible voices on the inside to keep sounding the alarm before these self-improving models slip entirely out of control.
Ultimately, we have reached a point where generalized warnings about the dangers of artificial intelligence have officially outlived their usefulness. If Silicon Valley researchers truly believe they are building a machine that will kill everyone, they need to start leaking the actual blueprints. Until these whistleblowers are willing to risk their non-disclosure agreements to expose specific corporate recklessness, the public remains entirely powerless to stop it.
