Before the meeting ends
An AI product needs to send your input to wherever its model runs. That input can include your prompt, a PDF or another attachment. The answer then travels back to you.
The model may run on infrastructure operated by a provider such as OpenAI or Anthropic, or on infrastructure run by another company. A product using a local model still runs that model on a physical machine. Local to the vendor does not necessarily mean local to your device.
The first question for the meeting is straightforward: where is our data going? Ask for an explanation of the route and the infrastructure involved before you agree to use the product.
- Ask where prompts, PDFs and attachments are sent for processing.
- An LLM runs on physical infrastructure, whether operated by a model provider or another company.
- Before the meeting ends, ask the vendor to explain the route your data takes.