The models behind the agent
LLM agent traces: what the model APIs say about training on them
Updated
Every agent step is a request to a model. If the model's provider trained on those requests, a buyer would be paying for something already used.
Ask about my agent logs 6 of the 19 agent platforms we read write that they will not train on what you send. Tell us where your traces sit and we will write back about yours, as a licence, a closing company's sale, or an introduction through your adviser.
The model APIs
OpenAI says data sent to its API is not used to train or improve its models unless you explicitly opt in, and that by default it does not use business data to train them. Anthropic says it will not use inputs or outputs from its commercial products to train its models by default.
Source: AI agent software data terms
Agent builders and coding agents
Writer says it will not use your materials to train its models, Lindy says neither it nor its model providers use your contributions for training, and Cursor's privacy policy says it does not use inputs or suggestions to train its models except in the cases it lists.
Source: AI agent software data terms
Where training is allowed
Galileo's terms give it the right to use customer data to train its algorithms internally, during and after the term.
Source: AI agent software data terms
On this site next
The rules on keeping the logs are on AI agent audit logs.