> #privacy

I wondered how long Claude Incognito chats are accessible. I noticed that even after the chat disappears from the web interface, you can still access it, e.g., if you still have a notification on your phone pertaining to an incognito chat. Anthropic reserves the right to store these for 30 days, but it seems you lose access to them somewhere a few hours after the last interaction to a few days (<5). I don’t know if that means they delete the chats early or just move them to a system that you can’t access as a user.

Overall, I don’t really feel safe enough in the self-regulatory sense to explore private topics in API chats, even ZDR I’m not sure would make me feel safe enough. And similar to how mass scanning of private communication undermines prevents free personal development, we need a true trustworthy ZDR-type or local AI that we can truly trust to not surveil us.

Finally someone went out to show it: every trace of information you leave in public can be scalably aggregated with LLMs to de-anonymize you. Every instance of “i work in field X” or “i’m too young for Y” can be combined to form a profile of you, and later linked to your name.

Every tweet, every comment on hackers news, it adds up and will eventually enable a linkage attack, where they have sufficient information to find a profile of yours with a name, e.g., on LinkedIn, or the specific project that you didn’t mention by name.

This is from a paper that dropped today on arxiv by Simon Lermen, Daniel Paleka et al. under supervision from Florian Tramèr

I somehow missed this, but OpenAI stated on October 22nd that they are no longer legally obliged to retain all outputs. The legal action by the New York Times led to a court order that compelled them to do so.