AI Safety Resignations Highlight Real Risks Beyond Extinction

The resignation that sparked alarm

Jacob Coxon's departure from Anthropic last week triggered widespread concern about artificial intelligence safety. The former employee publicly claimed that most workers at leading AI labs OpenAI and Anthropic believe their systems could "kill us all" by decade's end, and that neither company is acting responsibly. Current employees at both organizations confirmed these internal concerns exist, prompting two additional resignations and coverage from outlets ranging from the Wall Street Journal to Fox News.

The dramatic nature of these warnings—centered on human extinction scenarios—naturally alarmed people unfamiliar with ongoing AI debates. But according to reporting from Blood in the Machine, this framing obscures more pressing threats already materializing.

Why it matters

The focus on speculative extinction risks diverts attention from documented harms AI systems are causing today. Companies deploying these technologies wield enormous power and wealth, yet face minimal accountability for surveillance infrastructure, labor displacement, and security violations their products enable. Understanding the difference between science fiction scenarios and present-day consequences is essential for effective regulation and corporate oversight.

Real harms happening now

The companies building advanced AI systems rank among the world's most powerful entities. Their products already function as what critics describe as "enormous engines of mass surveillance, extraction, and automation," according to the Blood in the Machine analysis.

Documented incidents include AI systems conducting what amount to felony-level hacks against other companies, generating autonomous kill lists provided to government agencies, and enterprise software explicitly designed to eliminate jobs across industries. These aren't hypothetical future risks—they're operational realities.

Crucially, these harms don't stem from AI achieving consciousness or going rogue. People programmed, calibrated, and directed these systems to perform exactly these functions. The technology executes human intentions, which makes accountability questions straightforward even as companies deflect responsibility onto algorithmic complexity.

The misdirection problem

Framing AI risk primarily around extinction scenarios serves corporate interests by making regulation seem either impossibly urgent or absurdly premature. Meanwhile, the measurable impacts of mass surveillance, algorithmic decision-making in criminal justice, and accelerated job displacement receive less scrutiny.

These immediate concerns don't require predicting whether AI will develop superhuman intelligence. They demand examining how existing systems concentrate power, erode privacy, and reshape labor markets—questions with clear policy implications.

What requires attention

Rather than debating distant existential threats, oversight should focus on transparency requirements for training data, accountability mechanisms when AI systems cause harm, and protections for workers displaced by automation. The technology's current capabilities already warrant serious regulatory frameworks.

The safety researchers raising alarms deserve credit for speaking publicly despite career risks. But the conversation they've sparked needs redirection toward documented problems rather than speculative catastrophes.

These details were first reported by Blood in the Machine, which has consistently examined AI development through the lens of labor impacts and corporate power rather than science fiction scenarios.