alt Hacker NewsThis is a completely straightforward application of existing law on privilege. To maintain privilege, among other things, communications must be confidential.
Chatting with Claude breaks confidentiality: chats with Claude are subject to arbitrary inspection by Anthropic employees, not to mention the issue of model training.
You can use self-hosted LLMs without breaking privilege. And funny enough, law firms like Latham & Watkins are now buying Nvidia GPU clusters for this purpose.
Replies
> communications must be confidential.
Email and phone calls are considered confidential, even though it is possible for vendors to inspect the communication. I expect AI use will be tested legally, but once precedence has been established in court it will become a common practice.
That's Latham & Watkins, but I haven't heard of any others. I'm not convinced that for anyone smaller than them it is worth spending on HW. For adversarial works, you will always need the latest frontier model and those things are not available for most of the firms even for licensing, and it's much cheaper to use Amazon Bedrock or similar (which is already available). It's just not economical with the current trends. You won't be breaking any privilege even if the law firm is using a public cloud service. Maybe the law firm will have to use geofenced datacenters etc. based on deontology rules or national requirements... But in many jurisdictions, you have to use a law firm to keep LPP, that's the point.