A no-training setting or an enterprise agreement doesn't keep the prompt on your device. The provider can still decrypt it, and it can sit in retention logs. For client PII, PHI, financials or privileged material, "they won't train on it" is not "it never left the machine."
Locke works on your device, before the network call. You use ChatGPT as you do now, and the raw sensitive values stay behind.
The risk with ChatGPT
ChatGPT often runs on a personal account that corporate controls never reach. Pasted contracts, customer records and source code leave the device the instant you hit send, and a sent prompt can't be unsent.
How Locke helps
Locke detects sensitive values when you send a prompt and masks them on your device before anything reaches OpenAI. The Early Access desktop app covers the OpenAI API too. That includes the Codex CLI, your own scripts and anything else on the machine that calls api.openai.com without opening a browser tab.
Locke is free for one person
It runs on your device and masks sensitive data before a prompt is sent, on the AI tools it covers. Teams starts at $14.99 per seat a month.
