Hacker News
points by rglullis 4 days ago | 0 comments
If you *really* care, run your own inference. Everything else is just wishful thinking and/or convenience winning over self-sovereignty.
[–] addedGone 3 days ago
exactly, it's surprising to read this on HN with qualified individuals that assume that Cleartext chats and a contract is a form of technical security? :/
[–] RevEng 2 days ago
Breach of contract at an enterprise level is a corporate death sentence. Individuals might get screwed over but billion dollar corporations do not.

As an example, in the industry I work in, software license enforcement is quite easy to break if you want to. But you never would, because if your employer found out they would fire you on the spot. They don't want to violate their license agreements even if it meant free use because of the legal liability and the future loss of access to that product. Enterprises make a point of tracking their own license usage specifically to make sure they don't accidentally end up in violation. The systems are mostly about tracking and auditing for compliance, not about technical blockers to make it impossible.

[–] rglullis 1 day ago
The issue is that the contract can not be enforced. For your company to be penalized for violating the license agreement, first it needs to be detected by the license provider. If you are sending your code to OpenAI/Anthropic to get inference, how can one detect that it is not being copied by them and being used for training?