i am not heppy with any surveillance but the article mentioned two concrete cases.
-
samdesk a global crisis monitoting tool warned an executive of a demonstration that might put him in harms way. they used anthtopic's model to plan an alternate root.
-
a guy typed into claude that he had just bought an ak-47 and had the ceo in his sights. anthropic alerted the police but still wouldnt hand over the user's chats citing privacy policies. soooo thats zero surveillance.
the article states that based on some jobs postings, the article writer had concluded that anthropic is creating some pre crime unit. i cant say they aren't but the article can't say they are.