A researcher at the artificial intelligence company Anthropic has resigned and issued a public warning that the technology his industry is building could one day threaten the survival of humanity.

In departing, he gave up his stake in the company. That decision meant walking away from equity in one of the most valuable firms in the AI sector. He said the leading laboratories are moving too quickly, and he cast his departure as a way to speak plainly about the risks he sees.

In his own words, he said the pace of development amounts to "gambling with our lives," and warned that the most advanced systems, if left unchecked, could eventually "kill us all."

His warning lands at a moment when the technology has moved to the center of public and political life. The present period has been described as an "AI twilight zone," a stretch in which the tools are powerful and widely used but the rules remain unsettled.

That gap is visible in Washington, where Congress has not developed a framework to address the kind of extreme, long-horizon risks the researcher described, leaving federal policy largely undefined even as the systems grow more capable.

States have moved faster. In California, Governor Gavin Newsom signed a set of AI safety measures, legislation that drew support from major companies in the field, including Anthropic and OpenAI.

His warning is a prediction about what the technology could become, not a description of harm that has occurred. What is not in dispute is that a researcher chose to leave a leading laboratory, forgo his financial stake, and place his concerns on the record.

References

1. 1.Axios — researcher's resignation, forfeiture of equity, and quoted warnings about AI risk
2. 2.Axios — characterization of the current period as an "AI twilight zone"
3. 3.Public record — California Governor Gavin Newsom signing AI safety legislation and industry support from Anthropic and OpenAI
4. 4.Public record — absence of a federal framework in Congress addressing extreme long-horizon AI risks