The Only Thing Standing Between Humanity and AI Apocalypse Is … Claude?

Anthropic is locked in a paradox: Among the top AI companies, it is the most obsessed with security and the leader in researching how models can go wrong. But even though the security issues it identified have not yet been resolved, Anthropic is moving just as aggressively as its competitors toward the next, potentially more dangerous level of artificial intelligence. Its main mission is to figure out how to resolve that paradox.

Last month, Anthropic released two documents that both acknowledged the risks associated with that path and hinted at a path to avoid this paradox. A long-winded blog post by CEO Dario Amodei, “The Adolescence of Technology,” is nominally about “confronting and overcoming the risks of powerful AI,” but it spends more time on the former than the latter. Amodei cleverly describes the challenge as “difficult”, but his depiction of the risks of AI – made even more serious, he notes, by the high possibility that the technology will be misused by authoritarians – offers a contrast to his more upbeat previous proto-utopian essay “Machines of Loving Grace”.

That post talked about the nation of talent in the data center; The recent dispatch gives the example of “the dark sea of ​​infinity”. Paging Dante! Yet, after more than 20,000 mostly despairing words, Amodei finally strikes a note of optimism, saying that even in the darkest of circumstances, humanity has always prevailed.

The second document Anthropic published in January, “The Constitution of the Cloud,” focuses on how this move can be accomplished. The text is technically directed at one audience: the cloud itself (as well as future versions of the chatbot). It’s an intriguing document that reveals Anthropic’s vision of how the cloud, and perhaps its AI companions, will meet the world’s challenges. The bottom line: Anthropic plans to rely solely on the cloud to untangle its corporate Gordian knot.

Anthropic’s market differentiator has long been a technology called Constitutional AI. It is a process by which its models follow a set of principles that align its values ​​with overall human morality. The initial Cloud Constitution included several documents purporting to incorporate those values ​​– things like Sparrow (a set of anti-racism and anti-violence statements created by DeepMind), the Universal Declaration of Human Rights, and Apple’s Terms of Service (!). The updated version of 2026 is different: it is like a long sign outlining the moral framework that Cloud will follow, finding the best path of righteousness within himself.

Amanda Eskel, PhD in philosophy, who was the lead author of this amendment, explains that Anthropic’s approach is stronger than simply asking the cloud to follow a set of stated rules. “If people don’t follow the rules for no reason other than their survival, it’s often worse than if you understand why the rule is in place,” Askell explains. The constitution states that Claude is to exercise “independent judgment” when faced with situations that require balancing his mandate to help, protect, and integrity.

Here’s how Constitution puts it: “While we want the Cloud to be reasonable and rigorous when thinking clearly about ethics, we also want the Cloud to be intuitively sensitive to a variety of considerations and able to weigh these considerations quickly and intelligently in making live decisions.” intuitively There’s an obvious word selection here – the assumption seems to be that there’s more under the hood of Claude than just an algorithm picking the next word. The “cloud-stitution,” as one might call it, also expresses the hope that chatbots can “increasingly draw on their own knowledge and understanding.”

Intelligence? Sure, many people take advice from big language models, but it’s something else to claim that those algorithmic tools actually have the gravitas attached to such a term. Askell doesn’t hold back when I say this. “I think the cloud is definitely capable of a certain kind of intelligence,” she tells me.



<a href

Leave a Comment