Frequently asked questions

Saibre does not make ChatGPT compliant with anything. It removes patient identifiers on your computer before text or files reach the AI, and keeps a local log of what it protected. That is a technical control you can document in your risk analysis; your compliance program stays yours.

Any name or number you enroll is always masked (three on Free, unlimited on Premium and Professional). Premium also catches names and details in context without enrollment, and Professional adds the identifiers specific to healthcare, legal, tax, and financial work.

ChatGPT, Claude, Gemini, Copilot, and 20+ others: in the browser, in their desktop apps, in the terminal, and in files you upload. Phone apps are not covered; keep client work on your computer.

You need to be an administrator on your computer, and installation takes about two minutes. If someone else manages your machine, send them our security page — it is written for them.

No. Everything happens on your computer. There is no Saibre server in the path, and Saibre never receives your prompts, your files, or your secrets. Turn off Wi-Fi and it still masks.

Free protection for keys and your three enrolled keywords continues. Paid features stop at the end of the billing period. Every paid plan has a 14-day money-back guarantee.

Only AI sites pass through the local filter; banking, email, and your VPN go direct. A few apps that pin their certificates are blocked rather than leaked, and you can bypass them with one click.

No — and you don't have to take our word for it. Saibre is fully local: the domain allowlist is readable on your machine, logs contain detector names and counts (never content), and uninstalling removes everything — service, proxy settings, and certificate.

No. There is no Saibre server in the content path. This is architecture, not a privacy-policy promise.

No. Extensions protect one tab of one browser. Saibre protects the whole machine at the network level: every browser, desktop AI apps, IDE assistants like Copilot, and command-line tools — both directions, everywhere.

Attackers now target AI tools directly: hidden instructions in web pages, calendar invites, and emails can hijack an AI assistant without you clicking anything (look up EchoLeak, or the Gemini calendar-invite attack). The firewall inspects content headed for your AI tools and blocks known and obfuscated injection, jailbreak, and zero-click payloads before they land. It's part of the Premium plan.

Seatbelts exist for good drivers. 84% of AI users worry about their data — yet 30% paste confidential information anyway, and in 2025 around 100,000 shared ChatGPT conversations ended up indexed by Google.

Apps with certificate pinning can't be inspected, so their AI traffic is blocked rather than leaked — with a one-click bypass. Non-AI traffic is never touched. The full list of what Saibre adjusts is on the security page.

No, and be suspicious of anyone who says yes. Free masks secrets — API keys, tokens, private keys — plus 3 keywords you enroll. Premium adds personal identifiers, contextual masking, and injection blocking. Server-side AI connectors are out of reach of any local tool, and Saibre tells you when it isn't covering a path.

Coming soon. Saibre Desktop runs on Windows and macOS today.

Because your data never leaves your machine, we have no usage costs to pass on to you. Tools that scan prompts on their own servers meter by request or by gigabyte; there is nothing here to meter.

Self-serve billing portal, 14-day money-back guarantee, and Free stays free forever.