ONYRI Sanitize: Secure Your Data Before Sending It to AI

Research context and background

In the age of Artificial Intelligence, a single copy-paste action can inadvertently leak sensitive data outside your security perimeter. Whether it is a customer export, a support ticket, or an HR spreadsheet, pasting this information into an LLM sends it to external servers with no way to retrieve or control it once transmitted. Without protection, names, IBANs, API keys, and internal strategic decisions leave the company in the clear.

Benefits

ONYRI Sanitize is an in-browser anonymization engine designed to find private bits in your text and hide them before you send anything to an AI. The core philosophy is simple. Everything happens on your computer. Nothing is sent to a server. Key advantages include:

• 100% In-Browser: No installation is required and zero bytes are sent to a server.
• Local Analysis: All processing occurs locally on your machine.
• Reversible Tokens: Sensitive data is replaced with tokens that can be restored exactly in place once the AI response is received.

ONYRI strips 38 types of sensitive data across 6 families, covering three critical areas. For your contacts, it masks names, email addresses, phone numbers, IBANs, bank details, social security numbers, physical addresses, ID documents, and medical information. For your systems, it protects cloud provider keys, AI provider keys, payment processor keys, Slack tokens, inline passwords, and SSH keys. For your company, it hides internal project names, infrastructure URLs, meeting markers, salaries, company IDs, and client or partner names.

Use Cases

ONYRI operates entirely client-side through a three-step process. First, detectors spot sensitive data live as you paste your text. Second, each identified value is replaced with a reversible token. The cleaned text is then safe to send to your chosen LLM. Third, once you paste the model's response, the tokens turn back into your original data exactly in place, allowing you to review the full complete answer.

Users can add their own patterns using regex and assign them a dedicated color. The tool supports pasting text or dropping CSV and Excel files, applying tokenization column by column. For datasets exceeding 1,000 rows, processing moves to a dedicated worker to keep the interface smooth. Users can save combinations of detectors and rules to replay them in one click during future sessions. Teams can share rules and profiles across a workspace so the entire team applies the same anonymization policy.

Pricing

ONYRI offers a flexible pricing model with a 7-day Pro trial at 0 euros, charged only if the user continues after the trial. Plans include:

• Free: Unlimited text sections, 3 table files per month, 3 custom rules, full PII detectors, and 100% client-side processing.
• Pro: 22 dollars per month. This includes everything in Free, plus unlimited table sections, unlimited custom rules, multiple profiles, 90-day session history, and standard email support.
• Team: 57 dollars per user per month with a minimum of 3 seats. This adds shared workspaces up to 50 members, shared rules and profiles, roles and permissions, anonymized audit logs, and priority email support.
• Custom: This option is for unlimited seats, SSO, SAML, OIDC, SCIM provisioning, custom DPA, 99.9% SLA, a dedicated Customer Success Manager, and tailored onboarding.

Vibes

ONYRI Sanitize empowers individuals and organizations to leverage AI without compromising privacy. By ensuring that no sensitive data leaves your machine, it bridges the gap between powerful AI capabilities and strict data security requirements.

Additional Information

The tool guarantees zero data transmission. The server only serves the application. No text, no files, and no token-to-value mappings are ever sent to it. The mapping lives in memory and disappears on refresh. Nothing is persisted, ensuring no leaks. Detections are visible and editable before export, and users can define their own rules for specific work contexts.