Anthropic Rules 10% Extinction Estimate Is Not a Material Risk to Shareholders

SAN FRANCISCO — Anthropic's legal and investor relations teams concluded a three-hour review Tuesday after the company's Alignment Science lead publicly estimated the probability of AI-driven human extinction at greater than 10% within the decade, determining that the figure does not meet the threshold for formal shareholder disclosure because applicable securities law defines "material risk" as risk to the company, not risk posed by the company to the rest of the biological world.

The review also addressed Evan Hubinger's secondary disclosure — that Anthropic currently has no plan to solve alignment for superintelligence and is "not clearly on track" to develop one — and concluded this was already adequately captured by the phrase "forward-looking statements involve risks and uncertainties" on page 47 of the company's Series G investor documents.

A working group convened to address the plan gap has been given until Q2 to determine whether a timeline for developing a plan constitutes a plan. Three members declined to speculate on whether the group itself is on track.

Sources described Hubinger as "extremely principled" and "a genuine asset to the alignment function." The company confirmed he remains employed. A spokesperson said Anthropic "takes safety seriously," adding that the next model, Mythos 5.2, is scheduled for October and will include additional safety features not yet specified.

The company's November fundraising round, expected to value Anthropic at approximately $130 billion, will proceed as scheduled. Asked whether the extinction estimate would be discussed with prospective investors, the spokesperson said those conversations are private.

"Transparency is a core value," she added, "within the appropriate forums."