Security and data
The agents work on Claude models (Anthropic), so prompts and code go to the model provider under the terms of the account the work runs on.
In a pilot the work runs on your own API key, and model usage is billed to that key.
Our rule: client work runs on the client's API key or under commercial terms, never on consumer model plans.
Our rule: board and part designs on a critical path, that is flight, stop, power, ESC, the safety unit and the structure's load path, do not go to outside AI design tools or cloud design services; the factory's agents work on them on Claude models, under the account's terms, as above.
A board goes to the fab after a release act that a human performs.
In preparation: the guard that enforces this rule on the line.
In preparation: a separate system user for each line, the review in a container without network access, and write credentials kept out of the agents' sessions.
The line's ledger is hash-chained: editing or deleting a record breaks the chain, but the ledger is not tamper-resistant.
Report a security issue to office@promptcompany.ai with “security” in the subject.
This site sets no cookies and uses no analytics tools.