Savvai, Inc.

Helps you manage and document client-specific AI governance across ChatGPT, Claude, and Gemini

Your AI policy, put to work across ChatGPT, Claude, and Gemini. SavvaiGuard is for professionals using AI for their own work and client-related work. SavvaiGuard turns the hassle of implementing AI governance policies into a workflow that helps you more easily use AI. Set your firm's rules up once — and optionally each client's additional rules. Then SavvaiGuard applies those controls for you inside the AI tools you already use — ChatGPT, Claude, and Gemini. Move between models freely for whatever a task needs and get an integrated record of your AI use across all of them. SavvaiGuard also helps you carry out and document other governance activities called for by your policy, both overall and at the client level, through its Compliance Journal and professional AI review logging. Since SavvaiGuard is logging your firm and client AI interactions, a detailed governance report can be generated at any time. And SavvaiGuard is easy to set up and use for individual practitioners and small firms, requiring no IT resources. OPERATIONALIZE YOUR POLICY — SET IT UP ONCE Give SavvaiGuard your policy and it does the remembering. Import your written AI governance policy (PDF, Word, or paste it in) and SavvaiGuard reads it and turns it into working controls. Don't have a written policy yet? SavvaiGuard walks you through entering what your firm has decided — which AI tools you approve, how each kind of sensitive information should be handled, and what your firm treats as confidential — reviews your choices with you, and then produces a written Guard Governance Standard you can adopt, along with the configuration to run it. From then on, SavvaiGuard applies your configured Guard controls as you work — you don't have to keep them in your head. BUILT FOR MULTI-CLIENT WORK If you serve multiple clients — as a fractional CFO, attorney, consultant, or analyst — SavvaiGuard lets you layer each client's own requirements on top of your firm policy. So you never have to track which rule applies to which client. SavvaiGuard applies the firm and client controls you configured for that work, and keeps a separate, per-client record you can produce on demand. HOW IT WORKS SavvaiGuard reviews what you are about to send and applies the firm and client settings for sensitive information. If nothing in the submission requires Guard action, your work continues without interruption. Items set to Auto-sanitize are masked and sent automatically. If something requires your review, SavvaiGuard holds the submission in the review panel beside the AI tool so you can decide what to send. If something is set to Block, the submission is stopped. Each SavvaiGuard-reviewed submission is recorded with the applicable policy outcome or user decision. WORK ACROSS MODELS — ONE PLACE TO SEE IT Use whichever AI is best for the task and switch between them freely; SavvaiGuard applies the same policy across ChatGPT, Claude, and Gemini and keeps those interactions associated with the same client, whichever tool you used. Generate governance reports from the SavvaiGuard portal (guard.savvai.ai) and export to PDF, Word, CSV, or JSON — on demand, for any client, any date range. WHAT SAVVAIGUARD DETECTS SavvaiGuard combines Google Cloud Sensitive Data Protection with its own detectors, the client names you register, and the custom terms you add — so it recognizes the identifiers and confidential references that matter in your work, not just generic PII. DOCUMENT UPLOADS, GOVERNED BEFORE THEY ARE SENT SavvaiGuard reviews supported document uploads before releasing them to the AI — Word, PDF, Excel, PowerPoint, and CSV. If the document contains governed information, SavvaiGuard applies the settings for the selected firm and client. SavvaiGuard records the document outcome and the applicable policy decision. A downloadable detection report is also available for offline review or manual redaction. CONSISTENT TOKENS ACROSS A CHAT AND ITS DOCUMENTS One of SavvaiGuard's most powerful features is that when it masks a value, it keeps the same token for that value across all the messages in a chat — and across attached documents in the same chat. So when you ask a follow-up question, the AI still understands you're talking about the same thing, and its answers stay coherent. You get the protection of masking without losing the thread of your work. TWO MODES • Analysis mode — SavvaiGuard applies your configured Guard controls to governed content and handles it per your settings before submission, while preserving useful business context where your settings allow. • Research mode — sometimes the AI is only useful if it knows the company you are actually asking about. In Research mode, the registered company identity of the client you have tagged can reach the AI as written, so the AI can draw on outside knowledge about that company. Everything else in the same submission is still handled under your settings, and Research mode has no effect on your other registered clients. If you have set that client's company name to Block, or a custom term inside the name to Block, nothing is released. The interaction is recorded either way, with the mode noted. THE SAVVAIGUARD COMPLIANCE JOURNAL The Compliance Journal is where the governance work that happens outside the AI tools gets recorded alongside the work that happens inside them. Many policies call for more than in-session controls — AI-use training, client consent and disclosure, incidents, policy exceptions, and other human governance activities — and SavvaiGuard lets you log these as journal entries, so they live in the same per-client record as your AI interactions. Everything you record in the Journal feeds into your governance reports, so a single report can show both the automated activity SavvaiGuard captured, and the surrounding governance steps your policy requires. LOG PROFESSIONAL REVIEW OF AI OUTPUT Because many policies call for professional review of AI output before it's relied upon, SavvaiGuard lets you log a review — contemporaneously, with your notes — right in the SavvaiGuard panel while you're working. You decide which outputs warrant a formal review record; Guard captures the reviews you choose to log, so the affirmative step of a professional reviewing the output becomes part of the Journal rather than something tracked separately in email or a spreadsheet. PLANS • Advisor Trial — new accounts start with 14 days of full Advisor access, no credit card required, up to 100 interactions. Continue on Free after the trial or subscribe to Advisor any time. • Free — after the trial, continue with detect-and-warn protection: up to 50 interactions per month, 2 registered clients, SavvaiGuard detection panel, and record viewing. Masking, exports, client-specific settings, and governance reports are part of Advisor. • Advisor — $19/month — adds masking, per-client governance reports, exports (PDF, Word, CSV, JSON), client-specific settings, and use beyond the Free monthly interaction limit. PRIVACY AND SECURITY SavvaiGuard records detection results and your decisions — not the full text of your prompts or documents (policy documents you choose to store are the one intentional exception). Detection uses a combination of in-browser processing and secure backend services, and only detection metadata and decisions are kept in your record. GETTING STARTED Install the Chrome extension, sign in, and import or create your policy in the SavvaiGuard portal. SavvaiGuard is designed for self-service setup and works directly inside the supported AI tools without a custom integration project. FREQUENTLY ASKED QUESTIONS Does SavvaiGuard store my prompts or uploaded work documents? SavvaiGuard does not retain the full text of your prompts or uploaded work documents in AI Activity or governance records. It retains detection metadata and policy outcomes or user decisions. Separately, Policy Center can optionally store copies of your firm and client AI policy documents when you choose to enable that feature. Do I have to sanitize everything for SavvaiGuard to be useful? No. SavvaiGuard applies the Guard controls you configured — for some data that means masking, for other data it may mean review, allow, or block, and you decide at the point of submission. Whichever way you send, the interaction is recorded, so you always have a per-client account of how AI was used. What's the difference between Analysis and Research mode? Analysis mode is the normal working mode: SavvaiGuard applies your configured Guard controls to governed content and handles it per your settings before submission, while preserving useful business context where your settings allow. Research mode makes one exception, for the times the AI is only useful if it knows who you are asking about — a company's competitors, its public filings, who runs it. In Research mode the registered company identity of the client you have tagged can reach the AI as written, in what you type, what you paste, and documents you attach. Everything else is handled exactly as you configured it: Social Security numbers, dates of birth, people's names, account numbers, phone numbers, and your custom terms anywhere else in what you send. Other registered clients are not affected. If you have set that client's company name to Block, or a custom term inside the name to Block, nothing is released and the submission is stopped. Research mode requires a tagged client, cannot be switched on partway through a conversation SavvaiGuard has already reviewed, and resets when you start a new conversation or tag a different client. Both modes are recorded, with the mode noted. Do your company research and your client work in separate conversations — SavvaiGuard governs each submission as you send it, but it can't take back what a conversation has already been told. How does SavvaiGuard help with SOC 2, HIPAA, or NIST AI RMF? SavvaiGuard is AI governance infrastructure, not a certification. It helps you put governance controls into practice and produce a record that supports frameworks like SOC 2 and the NIST AI RMF — a per-client record and reports you can generate on demand. It does not by itself make your firm compliant. Does SavvaiGuard work with tools other than ChatGPT, Claude, and Gemini? SavvaiGuard currently supports ChatGPT, Claude, and Gemini in Chrome. Everything across the supported tools is recorded under the same client, without separate setup per model. Additional AI surfaces are under evaluation for future releases. Can I use SavvaiGuard on the free plans of ChatGPT, Claude, or Gemini? Yes. SavvaiGuard works with free and paid accounts on the supported AI platforms. For professional or client work, review the AI provider's current data-use, retention, and training settings and choose the account type that fits your organization's requirements. SavvaiGuard is developed by Savvai Inc. Support: supportrequest@savvai.ai Portal: Guard.savvai.ai Terms: savvai.ai/terms-of-service Privacy Policy: savvai.ai/privacy-policy