Optiq vs ChatGPT for confidentiality-sensitive professionals

OpenAI is transparent about how consumer ChatGPT works: conversations are stored in your account, and by default they are used to train OpenAI's models unless you turn on the opt-out. Even with training off, giving feedback on a response can send that whole conversation for training, and temporary chats are still kept for up to 30 days for abuse monitoring. None of this is hidden; it is simply the standard bargain of a consumer cloud product. The question for any team handling privileged, confidential, or patient information is whether that bargain fits the work.

Optiq was built so the question does not come up. Conversations live in an encrypted database on your machine, whose key sits in your operating system's keychain. Model calls are relayed through a confidential backend to private Venice models with zero data retention: nothing stored, nothing trained on, on any plan. When a chat never becomes a server-side record, there is nothing to opt out of and nothing to subpoena. There is more detail on our security page.

Credit where due: OpenAI has built out real compliance paths for regulated industries. For healthcare specifically, there are BAAs for the API, for sales-managed Enterprise and Edu accounts, a dedicated ChatGPT for Healthcare workspace, and a free ChatGPT for Clinicians tier where verified US clinicians can sign an in-product BAA. If your organization sets one of those up properly, that is a real compliance path, though it is healthcare-specific; ChatGPT does not offer an equivalent packaged compliance tier for law firms or financial advisory practices managing privileged or fiduciary data.

The catch is coverage. The BAA protects the seats and workspaces it covers, and nothing else. The clinician on the free Clinicians tier is covered; the practice manager pasting a patient email into their personal Plus account is not, and there is no equivalent tier at all for an associate at a law firm or an advisor at a wealth management practice pasting a client email into their personal account. OpenAI explicitly offers no BAA for consumer tiers or Business. Compliance that depends on every person being on the right tier with the right toggle is fragile. Optiq's answer is one architecture for every seat: local storage and zero retention are not a plan feature, they are how the product works.

ChatGPT's Work agent runs primarily in OpenAI's cloud: it operates on files and tasks on their infrastructure, with a desktop-local mode available on some plans. Optiq's agent runs on your machine, sandboxed by default, and anything risky, like writing a file or running a command, stops and waits for your approval. For file work involving sensitive material, where the agent runs is not an implementation detail. It decides whose infrastructure your documents pass through.