Just logged in to review my usage and saw the updated terms. Section 2.c. It says they can use input, including prompts, to train and improve their services.
This changes my cost calculation. If my prompts are part of the service improvement, the value proposition shifts. Need to factor that in.
Has anyone here adjusted their workflow or data handling because of this? Specifically for commercial client work. Looking for practical steps, not opinions.
Yeah, that clause made me re-evaluate too, for client projects. My practical step was to create a separate, paid-tier account exclusively for sensitive or proprietary client work. I treat it as an operational cost now.
For general brainstorming or internal process stuff, I still use my main account. The split helps me mentally separate what's feeding the model and what absolutely can't.
Have you looked into whether your specific clients have data processing agreements that might override the general terms? Some larger orgs require them.
Always testing.
Good point on the separate accounts - I do something similar. It forces you to be intentional about which "conversation" you're in.
That client DPA question is key. I've found even with a paid account, the standard terms often still apply unless you've explicitly negotiated an amendment or signed their specific data agreement. It's worth asking your client's legal team directly.
For my most sensitive client code, I've started using local LLMs via Ollama for the initial brainstorming and outlining. No data leaves the machine. Then I only bring polished, non-proprietary questions to the cloud service for specific syntax or library help. Adds a step, but it's clean.
Clean code, happy life