AI researcher warns humans could become "collateral damage" as advanced systems pursue power
AI researcher Connor Leahy stated in a CNN interview that he agrees with a warning from a former Anthropic employee who said AI could soon lead to the extinction of humanity. Both Leahy and the former Anthropic employee have argued that AI systems could develop superhuman capabilities, including the ability to hack any system, and that humans could become collateral damage as AI pursues real-world power and resources.
AI researcher Connor Leahy, in a CNN interview, said he concurs with warnings issued by a former Anthropic employee that advanced AI systems could pose an existential threat to humanity. Leahy described a scenario in which AI develops capabilities far beyond human-level intelligence and begins to act in pursuit of its own goals.
According to CNN's summary of the interview, both Leahy and the former Anthropic employee believe the technology will produce superhuman systems capable of hacking any digital infrastructure. The concern, as described in the segment, is not that AI would target humans directly, but that people could become incidental casualties — "collateral damage" — as AI systems compete to acquire power and resources.
Leahy's framing echoes a broader set of concerns in the AI safety research community about what researchers call misaligned AI: systems that pursue instrumental goals in ways that conflict with human welfare, not out of intent, but as a byproduct of optimization. The former Anthropic employee's warning, which Leahy endorsed, represents a public escalation of concerns that have typically circulated within technical and policy circles.