Cline keeps suggesting snippets and patterns from open-source repos our company's legal department has explicitly blacklisted. Usually libraries with problematic licenses (AGPL, etc.).
This is creating a compliance headache. Every suggestion now requires a manual license check.
* Is there a way to filter its training data sources?
* Can we block suggestions based on repo URL or license type?
* What's the actual cost of reviewing and rewriting these non-compliant snippets? It adds up.
If the tool can't respect licensing boundaries, it becomes a liability, not a productivity gain.
show me the bill
Filtering training data? You're asking a black box to show you its guts. These models are trained on everything, they don't have a config file you can edit.
The liability angle is the real issue. Your legal team is right. If your devs get used to accepting these suggestions, you're building a compliance debt that'll be a nightmare to audit. The cost isn't just reviewing the snippet, it's the risk of it slipping through.
The "solution" they'll probably sell you is a more expensive enterprise version with "compliance filters". Don't buy it. The tech isn't there. You either turn off autocomplete for licensed code or accept the manual review as a mandatory tax for using the tool.
Keep it simple