Privacy Policy

1. Who we are and what this policy covers

Microve, Inc. ("Microve," "we," "us," or "our") is a Delaware corporation operating in Nevada. This policy describes our handling of personal information on microve.ai, in the free Closed Circuit Professional Redaction app at redaction.microve.ai, and when you contact us or engage our professional services.

For a separately hosted or client-operated installation, the operator's configuration, privacy notice, and written agreement govern that installation. The public app's providers and retention practices should not be assumed to apply to another installation. A signed service agreement or data processing agreement may set additional obligations for a client engagement.

Contact us about privacy at support@microve.ai. Our Terms of Service explain the conditions for using our services.

2. Closed Circuit Professional Redaction

This section describes the free public app at redaction.microve.ai, offered for people who cannot afford subscriptions or only need one-time or occasional redaction. It helps you identify, review, remove, sanitize, and check sensitive information in one PDF, PNG, or JPEG at a time. It does not require an account or paid subscription.

What stays in your browser

The app reads your original file and performs document parsing, optical character recognition (OCR), manual marking, redaction, sanitization, and output verification in your browser. The original file, its filename, embedded metadata, comments, attachments, and page images are not uploaded by the current public workflow. Redacted files, verification certificates, and exported process logs are generated locally for you to download.

Your custom search-term list also stays in your browser. However, a term that appears in the visible page text can still be included in the text sent for AI detection. Information visible on a page is not excluded from detection merely because it also appears in metadata or a comment.

What is sent when you start detection

When you start AI detection, your browser sends visible page text, including text read from scanned pages or images by local OCR, to our server. Requests also contain page numbers, text positions, a randomly generated document-session identifier, and detection context such as previously identified values and questions about ambiguous matches. Our server sends the page text and relevant detection context to the AI service and returns suggested findings to your browser.

This happens before you review and apply redactions. The transmitted text can include names, contact details, government identifiers, financial information, health information, information about children, or other confidential information present in your document. It can concern you or other people. Choosing a file alone does not start AI detection; starting detection initiates these requests.

AI requests are made page by page, with some findings carried forward as context. The current public workflow does not send page images to a vision model. Output verification runs locally and does not send the finished document to an AI reviewer.

Providers involved in detection

• Railway hosts the redaction application server in the United States and handles the encrypted connection from your browser. Its infrastructure receives the detection requests and ordinary connection information.
• OpenRouter routes the detection text to an eligible model provider and returns the response.
• Microsoft Azure is the currently configured model host for the OpenAI GPT-4.1 mini model reached through OpenRouter. The public app requires zero-data-retention routing and disallows provider data collection on each detection request.

These routing restrictions select endpoints that OpenRouter classifies as meeting the requested data policy; the text is still processed by external services. OpenRouter's definition allows temporary in-memory prompt caching. OpenRouter also records request metadata, such as token counts and latency, and describes anonymous categorization of sampled prompts. Its own prompt and response storage is opt-in, separate from provider routing. See OpenRouter's zero-data-retention explanation and data collection disclosure.

We do not use your documents or detection text to train, fine-tune, evaluate, or improve AI models, or build datasets for those purposes. Zero-data-retention routing does not mean that no information leaves your device or that service providers retain no operational or billing metadata.

Application logs and abuse prevention

Our redaction server does not create a document archive, database of page text, or saved history of model responses. It processes detection content in memory for the request. Its application logs record timestamps, request types, status and error codes, durations, byte and token counts, and the random document-session identifier. They exclude document text, filenames, model answers, and raw IP addresses.

For usage limits, the server receives your IP address and derives a temporary keyed identifier for your network. It holds that identifier and usage counters in memory; it does not keep the raw address in the rate-limit records. These records are cleared on a server restart or removed as entries become inactive and their usage allowance resets. People on the same network can share a limit.

Railway and other infrastructure providers separately handle connection and security metadata, which may include IP addresses and request details. The limits on our application's logs are not a claim that hosting providers collect no technical information. See retention below.

We may use available operational records to investigate abuse, enforce limits, restrict access, and protect the service. Relevant records may be preserved or disclosed where legally appropriate for an investigation or legal claim. This does not mean we keep a routine archive of document text. We will update this notice if additional abuse controls materially change our collection or use of personal information.

Your session and downloaded files

The app keeps document work in the current tab's memory, rather than an account, cookie, or persistent document store. Starting over, refreshing, or closing the tab discards the working session; we cannot restore it. This is not a promise of forensic erasure from your device's memory, swap, browser, or operating-system backups.

Originals and downloads on your device remain under your control, including any copies made by your browser, backup software, or cloud-sync service. A verification certificate describes checks on the generated output; it does not certify deletion of the original or every other copy.

Confidential and regulated information

Only process information you are authorized to disclose to the providers described above. You are responsible for permissions relating to other people's information and for your organization's confidentiality, professional, and legal obligations.

No HIPAA business associate agreement covers the free public app. Do not use it to process protected health information where a business associate agreement is required. Selecting a HIPAA or court-rule preset does not make the service compliant or establish that a document is legally de-identified. Contact us about a separately agreed deployment if you need particular contracts, locations, or processing restrictions. See the redaction terms.

3. Website visits, inquiries, and other services

Information you provide

When you email us or discuss a project, we receive the information you choose to provide, such as your name, email address, company, telephone number, message, and attachments. Our current contact page directs you to email. Client projects may involve additional information described in the applicable agreement. Do not email original sensitive documents merely to report a redaction issue; a description or synthetic example is usually sufficient.

Connection information

Our website host, Vercel, and services delivering website assets receive ordinary request information, such as an IP address, browser information, requested URL, time, response status, and any referrer sent by your browser. Website pages use Google Fonts and the Tailwind styling CDN, so those providers receive requests when their assets load. These website services are separate from the redaction app, which serves its fonts and processing assets from its own origin.

4. Cookies, browser storage, and tracking choices

Redaction app: the current app does not set application cookies, run advertising or analytics trackers, or save your document session in local storage or a browser database. Your browser may cache static application files such as scripts, fonts, and OCR assets.

Main website: site pages show a cookie banner with Accept, Preferences, and Reject. Preferences opens a panel where necessary cookies stay on and analytics can be turned on or off. A footer link labeled Cookie settings opens that preferences panel. The preference cookie is named microve_cookie_consent. It stores a version number, that necessary cookies are in use, whether analytics is accepted or rejected, and the time of the choice. It does not store a marketing choice. The cookie lasts 365 days. It uses path=/ and SameSite=Lax, and it includes Secure when the page is served over HTTPS.

Google Analytics is the only optional tracker. It loads only after you choose Accept, or after you save preferences with analytics on. Before any Google tag configuration, Google Consent Mode sets analytics storage, ad storage, ad user data, and ad personalization to denied. Accept, or saving preferences with analytics on, changes only analytics storage to granted. Ad storage, ad user data, and ad personalization stay denied. We do not use advertising or marketing cookies. While Google Analytics is loaded, Google can receive page and connection information and can set analytics cookies such as _ga. Rejecting analytics, saving preferences with analytics off, or not choosing yet keeps that script from loading. A later visit loads Google Analytics only when the stored choice allows analytics and the browser is not sending Global Privacy Control. Fonts, styling, and other files the page needs to display still load, and those providers still receive ordinary connection information.

When analytics is allowed, the site can send two events. consultation_click is sent when you click a link or button to the contact page or a consultation booking. generate_lead is sent only after a contact form reports success, not when the button is first clicked. Each event includes the page path and the visible label of that link or submit button. The events do not include what you typed into a form, and they do not include your name, email address, or phone number. If analytics is rejected, or Global Privacy Control is on, these events are not sent.

Cookie settings opens the preferences panel, where you can change or withdraw consent. Saving with analytics off sets analytics to denied, stops analytics, and deletes _ga and _ga_* cookies on the current host and tries to delete them on .microve.ai. If your browser sends Global Privacy Control, we treat that as a rejection: we do not prompt you to allow analytics, we store the rejection in the preference cookie, and Google Analytics does not load. The preferences panel keeps analytics off while that signal is on, and saving preferences does not load analytics. The site does not change its behavior for a Do Not Track signal. Browser settings also let you block or delete cookies. We do not sell personal information or share it for cross-context behavioral advertising.

5. How we use information

• Provide the requested service, including returning redaction suggestions and creating outputs in your browser.
• Respond to inquiries, provide support, and manage client relationships and contracts.
• Keep services available, diagnose failures, enforce usage limits, and prevent abuse using operational information.
• Meet legal obligations and establish, exercise, or defend legal claims.

We use client content to deliver the requested service, not to train or improve AI models. Operational counts and performance information may help us improve reliability. We do not create advertising profiles from redaction text.

Where applicable data protection law requires a lawful basis for our own processing, we rely on performing a contract or taking steps you request before a contract, legitimate interests in operating and securing our business and responding to inquiries, compliance with legal obligations, and consent where required for optional technologies or communications. When we process information on an organization's instructions, that organization is responsible for its lawful basis and any additional conditions for sensitive information; the applicable processing agreement sets out our respective responsibilities.

6. Who receives information

We disclose information as needed to provide the services described here. The current providers and their roles include:

• Railway: redaction hosting and network infrastructure. Privacy policy.
• OpenRouter and Microsoft Azure: redaction model routing and inference, as described in section 2. OpenRouter privacy policy; Microsoft privacy statement.
• Vercel: website hosting. Privacy policy.
• Google: business email, website fonts, and the optional script described above. Privacy policy.
• Tailwind CDN: delivery of website styling code; it receives the connection information needed to deliver that code.

For client projects, additional providers may be identified in the relevant service or data processing agreement. Provider policies describe their own processing; they do not replace Microve's commitments in this policy.

We may also disclose information we actually hold to authorized personnel and professional advisers who need it, when legally required, to protect rights and safety, or in connection with a merger, financing, or business transfer subject to appropriate protections. These circumstances do not create a document archive where none exists.

7. How long information is kept

• Redaction document content: the working session is held in browser memory, and detection content is processed in server memory for the request. Our application does not intentionally retain those documents, page texts, or model responses after processing.
• Redaction usage limits: keyed network identifiers and counters remain in server memory until reset, expiry through inactive-entry cleanup, eviction, or restart.
• Redaction application logs: Railway currently provides a seven-day log window on the service's hosting plan. This is the accessible log window, not a guarantee that all provider-held information is erased on day seven. Provider security, billing, and other operational records follow their own policies. Railway log documentation.
• AI service metadata: OpenRouter and model providers may retain operational, billing, or security metadata under their policies. The zero-data-retention restriction on inference content is separate from those records.
• Inquiries and support: we retain correspondence for as long as needed to respond and manage the relationship, with longer retention where necessary for contracts, legal obligations, or disputes.
• Client project data: retention, return, and deletion follow the applicable agreement and legal requirements.
• Website records: preference-cookie duration is described in section 4. Hosting logs are subject to the hosting service's retention settings and policies.

We cannot recover or provide a copy of a redaction session we have not stored. Deleting a browser session does not delete your originals, downloads, support emails, or records held by providers for their own operational purposes.

8. International processing

Microve operates in the United States. The public redaction server is hosted in the United States, and our providers may process information in other countries. The free public app does not offer a country-specific processing or data-residency commitment.

If your organization requires a data processing agreement, contractual transfer safeguards, or a particular processing location, contact us before sending information. Using the public app or accepting these terms does not itself create a signed data processing agreement, Standard Contractual Clauses, or a HIPAA business associate agreement. Where applicable law requires transfer safeguards, those requirements must be met; this notice does not waive them.

9. Your choices and privacy rights

You can choose not to start AI detection, stop using a service, clear your browser session, manage website cookies, and ask us about information we hold. Depending on the law that applies, you may have rights to access, correct, delete, or receive a portable copy of personal information; restrict or object to processing; withdraw consent; or opt out of certain processing. We do not sell personal information or share it for cross-context behavioral advertising.

Send requests to support@microve.ai. We may need proportionate information to verify your identity or an authorized agent's authority. We respond within the period required by applicable law. Legal exceptions may apply, and we will not discriminate against you for exercising applicable privacy rights. Where available, you can appeal a decision by replying to our response.

If another organization supplied your information through our services, contact that organization as well; we can assist with requests as required by our obligations. Please do not send an entire confidential document to identify a request. You may also complain to the relevant data protection authority, including in the EEA, United Kingdom, or Switzerland where applicable.

10. Security

We use measures appropriate to the services, including HTTPS for public connections, restricted application requests, limited redaction logging, and controls on access to service infrastructure. No online service or device is completely secure. The app's privacy features and abuse controls do not guarantee that all misuse, unauthorized access, or disclosure can be prevented. These limitations do not remove our obligations under applicable law or the commitments in this policy.

The app also cannot guarantee that every sensitive item is found or removed from an output. Review the document before sharing it. You are responsible for the security of your device, browser extensions, originals, and downloaded outputs. See the redaction terms for review responsibilities and the limits of verification.

11. Children

Our services are intended for adults and organizations, not for children under 13 to use. Documents submitted by authorized adults may contain information about children, which is processed as described above. If you believe a child has provided information directly to us without appropriate authorization, contact support@microve.ai so we can investigate and take appropriate action.

12. Changes and contact

We will update this policy and its date when our practices change and provide additional notice or obtain consent when required by law. For material changes to redaction processing, we will update the app's disclosure and this policy. This notice does not authorize new uses of previously collected information where further permission is required.

For questions or privacy requests: Microve, Inc., a Delaware corporation operating in Nevada, support@microve.ai.