I installed buzzy AI model Jev in a contained repo this morning to test out the hype they’re trying to manufacture on Reddit. I’m not a hater of the hustle from their growth team. It’s great for isolating tool calls and lowering the cost of tokens with whatever frontier model you’re using.
BUT it definitely opens up a conversation that AI startups need to have earlier than their counterparts of ten years ago. Closed AI model businesses are a black box. Unlike open source software or to a lesser degree open weight models, you’re not really sure how this data is being fed into their systems.
With that in mind…
If you’re building an AI company and are a black box and have broad interpretation in your terms of service into how your data is used and not used, you need to have clear limitations in how data is preserved (what, for how long) and how it’s removed (what causes permanent data deletion and for how long).
Customer Personal Data is retained “for as long as necessary taking into account the purpose of the Processing.”
and the privacy policy…
“as long as reasonably necessary to provide you with the Services, or otherwise in support of our business or commercial purposes.”
The tool is awesome and noticeably sped up Claude Code in tests I ran, but AI’s a work tool. I can’t use Jev for work (read: pay for it) without clear communication on client data.
The risks for companies that hold data in perpetuity by default is sky high and the lack of consumer/enterprise data protections in the U.S. following the financial distress of a company are not great.
Leave a Reply