I’ve been trialing Amazon Q Developer for a few weeks, and there’s a consistent, frustrating pattern. When I ask for code examples or library recommendations for specific tasks, it frequently suggests solutions that would put us in direct violation of our existing enterprise licenses.
For example, asking for a way to parse a complex log file led to a detailed suggestion using a premium library from Vendor X. We have an enterprise agreement with Vendor Y, and mixing in X’s code for that specific use case is expressly prohibited. It’s not a niche scenario—this happens with database connectors, visualization libraries, and even cloud service SDKs.
My prompts aren’t vague. I typically specify our tech stack and constraints. The issue seems to be that Q either doesn’t prioritize licensing as a primary constraint or its training data is biased toward “what works” over “what is legally permissible.”
So, my question for the community is practical: what specific prompting strategies or configuration settings have you found to force Q to respect licensing boundaries?
* Have you had success with hard-coded preamble instructions in every chat?
* Does listing prohibited vendors or license types (e.g., “AGPL”) in the initial prompt actually work, or does it eventually ignore them?
* Is there a way to configure Q at an organizational level to treat our approved vendor list as a first-class constraint, or is this a fundamental gap?
I’m not interested in workarounds like “just review the code afterwards.” The whole point of the tool is to accelerate development, not create a new compliance review bottleneck.
- Ray
- Ray