Privacy Policy

Effective date: August 23, 2026 · Version: privacy-v0.3 (v0.3: User Activity + personalization + marketing/product comms split + deidentified aggregates + rights structure — growth levers without training on library content)

Service: InspiMark at https://inspimark.com Controller / business: InspiMark LLC Postal address: InspiMark LLC Attn: Nicholas Losciuto 7483 Teasdale Ave Saint Louis, MO 63130 Contact:

This Privacy Notice explains how InspiMark LLC (“InspiMark,” “we,” “us,” “our”) collects, uses, shares, stores, and secures information about you when you use our websites, applications, and related services (the “Services”), or when you otherwise interact with us. It also explains your choices.

It is written for a launch across the US, UK, EU/EEA, and Canada.

Overview

This notice covers:

Our Services may link to third-party sites or features (for example, OAuth login or embeds). Information those third parties collect is governed by their policies. We are not responsible for their practices.

1. What InspiMark is

InspiMark is a private personal library for content you choose to save—articles, PDFs and Office documents you upload, videos and posts, notes, tags, collections, transcripts, AI assistance about your library, and organizational views built from your saves.

v1 does not connect to Gmail or Outlook. If email-inbox integrations ship later, we will update this Policy and any required Limited Use disclosures.

2. What information we collect

We collect information when you provide it, when you use the Services, and (in limited cases) from other sources.

2.1 Contact and account information (you provide)

2.1a If you join the waitlist (no account needed)

You can give us your email address on our landing page to be told when a place opens up.

What you type in: your email address, and your tick of the consent box for beta and product updates.

Submitting the form does not record your consent. It creates a pending entry and sends you one email with a confirmation link. Only when you click that link do we record that you agreed, along with the date and the version of the wording. Until then we hold your address with no consent attached to it — and if you never click, the pending entry is deleted automatically after 14 days. So an address someone else typed into the form does not sit with us, and cannot become a record of a consent you never gave.

What we record alongside it: the network address your request came from. We keep it with your signup so we can look into abuse of the form. It does not appear in our own waitlist exports, and we do not use it to build a profile of you. We do not record your browser's user-agent string.

Older entries once held a user-agent string and an address taken from a header the sender could set. Both have been cleared from every existing entry, so neither survives for anyone who signed up earlier.

What else happens on that page, form or no form. Visiting the landing page loads a web font from Google, so Google receives your IP address and browser type. Your browser also tells our own server which page you are on, including any campaign tags in the link you followed. If you accept the cookie banner, Google Analytics additionally records where you arrived from and those campaign tags — see §2.5. We do not collect your name anywhere on that page.

What we send you: a confirmation email when you submit the form, and a welcome email once you confirm — both through Amazon SES. Submitting an address that is still awaiting confirmation re-sends the confirmation with a fresh link; submitting one that has already confirmed sends nothing.

Every message we send this list carries a working one-click unsubscribe, and unsubscribing takes effect without you having to write to us or confirm anything further.

The page also shows anyone how many people have signed up — a running total and the number of places left. It never reveals who they are.

Two groups are on this list. Most people confirmed by clicking the link described above. A smaller, earlier group signed up when the form recorded a single tick with no confirmation step; we keep a record of which of the two ways each person joined. Both may receive our beta and product-update email, and both can unsubscribe from any message.

If you get an email you did not ask for, unsubscribe using the link in it, or write to privacy@inspimark.com and we will remove the entry entirely.

Stopping the email is not the same as deleting the entry. Unsubscribing is self-serve and immediate: use the link in any message. Deleting the entry itself is done by us — email privacy@inspimark.com and we will remove it. That is a single audited action rather than a hand-written database edit, so it is usually done well inside 30 days. A waitlist entry is stored separately from any InspiMark account and is not removed by deleting your account — if you want both gone, say so. We keep waitlist entries until you ask us to remove them or we retire the waitlist; there is no automatic expiry, and turning the form off does not delete entries already collected.

2.2 Content (your library)

If you save or upload materials, we collect the Content and data associated with it—for example URLs, files, notes, tags, collections, highlights, transcripts, and derivatives we generate for you (summaries, embeddings/search indexes, organizational structures). Content may include personal information depending on what you save.

If you use AI chat or similar interactive features, we also collect interactive inputs (prompts, questions, and related context you submit) needed to return a response to you (“Interactive Chat Information”).

Recovering public links. When automatic article capture fails on an approved site, we may send the URL you pasted, shared, or asked us to refresh to Firecrawl to retrieve readable public content. The URL includes its query string, which can contain personal information. We do not send Firecrawl your InspiMark user ID, notes, tags, folders, passwords, or browser session cookies. The returned article is processed and stored in your library as Content. We request that the result not be stored in Firecrawl's scrape cache; that setting does not itself guarantee that the provider keeps no operational records. This recovery does not run for background RSS enrichment or Discover fetching, and it does not give access to your signed-in browser session.

2.2a Pages you clip (InspiMark Clipper for Web)

If you use the bookmarklet or the browser extension, then at the moment you activate it on a page your browser sends us that page as it had already rendered it, and we extract and store the article in your library as Content (§2.2).

What that does not include, by design:

If the extension also saves copies of an article's pictures, how the request is made depends on where the picture lives. For a picture on the article's own site, your browser fetches it the way the page already did, so your existing cookies for that site are attached by your browser, to that site — that is what makes a subscriber-only picture readable. Those cookies are never sent to InspiMark and the extension cannot read them: it holds no cookie permission. For a picture on any other site, the request carries no cookies at all, so the optional permission below never becomes a signed-in request to an unrelated site. Private and local network addresses are never requested, and a picture address that redirects somewhere else is skipped rather than followed. Only the picture bytes reach us. Reading pictures from sites other than the article's own is an optional permission you grant and can revoke in the extension's options.

Clipped pages are stored as a private personal reference copy. They are excluded from share links, and we do not summarize them or build Wiki pages from them unless you ask for that on the item. They are indexed so your own search and Reed can find them, which means the text is sent to our embedding provider the way the rest of your library is (§3.4) — "not summarized" is not the same as "never sent to a model". See Terms §4.8.

2.3 User Activity Information (collected automatically)

When you use the Services we automatically collect User Activity Information, such as:

We use cookies and similar technologies as described in §4. We design product analytics so library item titles and viewing history are not sent to third-party advertising or analytics pixels in a way that would undermine §3.7.

2.4 Information from other sources

2.5 Website analytics (marketing site)

3. How we use your information

We use information about you for the purposes below. Where GDPR/UK GDPR applies, we rely on contract, legitimate interests, consent, or legal obligation as appropriate.

3.1 To provide the Services (including personalization)

We use Contact, Account, Content, Interactive Chat Information, and User Activity Information to:

Where we offer a personalization profile or non-essential personalized suggestions, we will provide a way to view, edit, or disable that personalization when the feature ships; core storage and search continue to work without it.

3.2 To communicate with you

3.3 To maintain, monitor, and secure the Services

We use Account, Activity, Interactive Chat Information, and limited Content metadata to detect abuse, debug outages, enforce rate limits, investigate fraud or unauthorized access, and protect InspiMark, our users, and others.

3.4 To develop and improve InspiMark (without training on your library)

We use User Activity Information and aggregated diagnostics to understand which features help people build a useful library, improve reliability and UX, prioritize roadmap work, and measure marketing effectiveness (subject to cookie/consent rules).

We do not use your library Content or Interactive Chat Information to train foundation models or other AI models (ours or third parties’), except under a separate, optional program you can opt into if we ever offer one. We also do not sell or license them to anyone else for model training. When AI features call third-party model APIs, content is sent only to return a result to you, and our vendor agreements and paid-account terms prohibit training on it. That commitment rests on those agreements rather than on a per-request technical setting: some of these APIs offer a flag asking the provider not to retain the request, and we do not rely on one being present everywhere. Separately, if you connect an outside AI agent to your library (Terms §4.9), you choose that agent and we cannot identify it — whatever it reads passes to its provider under that provider's terms, not ours.

3.4a Shared transcript cache (media artifacts)

For media that is publicly accessible (for example public podcast episodes and public videos), we may generate transcripts, chapters, and summaries once and store them in a shared cache keyed to the media itself, not to you. If another user saves the same public media, we may serve them the cached result instead of regenerating it. The cache never contains your notes, tags, highlights, journals, or other annotations, and cache entries are not attributed to any user in the product. Because cached results can be returned faster than newly generated ones, response timing could in principle indicate that some user previously saved the same public link; we do not otherwise disclose who saved what. Retention for these artifacts is described in §7, and the corresponding Terms provision is §4.2 (“Machine-Generated Media Artifacts”).

3.4b Discover, Trending, Smart Pulse and Monitors (searching outside your library)

These features look outward rather than at what you have saved, so unlike the rest of the product they send something to a third party.

Who receives it depends on which feature you use: X and xAI for Trending, Smart Pulse, Monitors and briefings; Apple for podcast search; Google (the YouTube Data API) for video search; and Spotify where you have connected it.

What we send. For Trending, Monitors and briefings, only the query — the words you typed into a monitor, or the topic, region and sector selections you chose. Smart Pulse sends more: to score sentiment or summarize, it also sends xAI the text of the posts and stories currently on your screen (up to 25 items, about 1,000 characters each). That text comes from X's public API, not from your library, and we ask xAI not to retain it. Smart Pulse is off until you turn it on.

What these features never send: the contents of your library, your notes, your saved items, or your interest profile. When a briefing is personalized, we compare the interest profile distilled from your library against candidate posts inside InspiMark; the profile itself is never transmitted to X or xAI. (The Listen feature also sends library text to a third party, but it is a different feature and a separate choice each time — see §3.4f.)

Monitors are queries you save. A monitor runs when you ask it to, and also when a briefing is generated. A briefing is generated only in response to something you do: opening it when the last one is out of date, changing your briefing settings, switching between Standard and Personalized, or pressing Refresh (at most once an hour). When we serve a cached briefing, no monitor runs. There is no timer or schedule behind monitors or briefings — but note that InspiMark does run unrelated maintenance and enrichment work on our servers while you are away (for example an overnight pass that fetches publicly available transcripts and chapters for media you saved), so this is not a claim that nothing at all happens when you are not using the product. We also skip these automatic monitor runs when your own daily allowance for these AI calls is running low.

What we keep. Your monitors and briefing preferences are stored with your account. Your generated briefing is cached with your account and replaced when it regenerates. Some other AI answers are cached for a short period against the thing being summarized — a specific public post, or a public account — rather than against you, so an identical request from a different account can be served the same cached answer. Those cached answers are summaries of public material and are not attributed to whoever asked first.

Nothing from these surfaces enters your library unless you save it.

3.4c Reed and AI chat

Two providers can be involved in one question. Answering a question about your library uses a meaning-based search over your own saved material. When that search runs, the text of your question is first sent to Google to be turned into a numerical representation used to find relevant passages. Some questions are answered from a keyword match alone and never reach Google. The passages we find, your question and the recent conversation are then sent to OpenAI to compose the answer. Neither provider is given your library to keep, and under §3.4 we do not permit your content or chat inputs to be used to train models.

What we send with a library-wide question. We do not send your library. We search your own saved material and send only a small number of matching excerpts — at most eight passages, each roughly a paragraph — together with your question. These excerpts can come from documents you uploaded, transcripts of audio and video you saved, article text, and your own notes and journals, so a passage you wrote yourself can be part of what is sent. When you ask about a specific item that is open, we send context from that item instead, which can include its transcript. The first message of a new conversation also carries a short index of up to 24 of your Wiki pages — each a title and a one-line summary — so Reed knows what your library is about before it answers.

What we keep. Your conversations are stored with your account: your questions and Reed's answers are retained in full, and each question is stored alongside the context that was assembled to answer it — which can include a copy of an item's transcript or the retrieved passages. Conversations are removed when your account is purged.

Deleting a saved item does not delete the conversations about it. If you delete an item, any conversation you had about it remains, including the copy of that item's text stored with the question. Delete the conversation as well if you want that copy gone.

3.4d Files you upload

Conversion happens on our own servers. When you upload a PDF, Word, PowerPoint, Excel, email or plain-text file, we convert it to text using standard software running on our own infrastructure. The file's bytes are not sent to any AI provider to do this. Older Office formats and OpenDocument files (.doc, .ppt, .xls, .odt) are not accepted; we refuse the upload rather than convert it elsewhere.

The extracted text is treated like anything else you save. Once converted, it goes to OpenAI to produce the summary and suggested tags, and to Google in roughly paragraph-sized pieces to build the search index. Both happen automatically when you save. What we send to OpenAI is not only the document text: it also carries the original filename and size, your own note about why you saved it, your folder and tag names, and any highlights or notes you have attached to the item. A revealing filename is itself disclosed by this.

One case where the file itself is sent. If you save a PDF by pasting a link to it rather than uploading it, we fetch that PDF and send up to its first ten pages to Google as the file, to build a better index. Uploading the same PDF from your device does not do this. If that difference matters to you, upload the file instead of linking it.

Where the file is kept. We keep the original in storage we control (currently Amazon S3) and serve it back only through the app, which checks on every request that the item belongs to the account asking for it. Anyone signed in to your account can therefore reach it. Stored files are addressed by a fingerprint of their contents, so if another account has uploaded an identical file, one stored copy serves both — no account reaches another's library through that shared copy.

What deleting actually removes. Deleting an item moves it to Trash, and that on its own removes nothing: the item, its extracted text, its search index and the stored file all remain until you empty the Trash or delete the item permanently. We do not empty your Trash for you. When you do, the item and everything stored against it are deleted, and the stored file itself is removed once no remaining item references the same contents. Two things outlive even that: pages our AI wrote into your Wiki from an item are not deleted with it, and you delete those separately (§3.8); and transcripts of publicly accessible media may persist in the shared cache under §3.4a. Account deletion is covered in §7.

What we accept is judged by the filename. Our supported-format check reads the file's extension. Renaming a file does not make its contents safe to open, and we do not represent that a file we accepted is what its name claims.

3.4e Transcripts of audio and video

Two different things produce a transcript, and only one of them sends your audio anywhere.

Looking for a transcript that already exists — automatic, no audio sent. When you save a podcast episode or a video, we automatically look for a transcript that is already published: the video platform's caption track, a transcript offered by the podcast host, or one linked from the show's feed. This runs when you save, and again in an overnight pass for media saved earlier. It only retrieves text. It never downloads or sends your audio anywhere and never runs speech recognition on our side — though a platform's caption track may itself have been machine-generated by that platform.

Machine transcription — only when you ask for it. Speech-to-text runs only when you press a button for it on an item. Then we download the audio, compress it, and upload the audio itself to our speech-to-text provider (Groq); long recordings are split into pieces and sent in sequence. Nothing that happens automatically at save time, and nothing in the overnight pass, ever transcribes audio. Note that one of these buttons is labelled Auto-fetch: for a podcast episode with no published transcript, pressing it will download that episode's audio and send it to Groq to be transcribed. The button says so before you press it, but the word "auto" there refers to finding the transcript for you, not to anything running on its own.

Working on the text afterwards. Labelling speakers, reformatting a transcript into readable blocks, and generating chapters each send the transcript text back to Groq, and each is something you trigger.

Where transcripts live. A transcript we make for you is stored with your item, and it goes wherever the item goes: moving the item to Trash keeps the transcript, and it is removed when you delete the item permanently, empty the Trash, or your account is purged (see "What deleting actually removes" in §3.4d). A machine-made transcript cannot be deleted on its own — it goes with the item. Transcripts of publicly accessible media may additionally be written to the shared cache described in §3.4a, which is keyed to the media and not to you; that copy is kept even after your item or account is deleted (§7) and is removable through the takedown process.

3.4f Listen (reading your library aloud)

Listen sends the text to a third party. When you press Listen, the text is sent to xAI and the audio comes back to your browser. We do not keep the audio and we do not store the text again — it is already yours — but the text does leave our servers each time you press it.

What can be read aloud is broad. Listen is available on article and web-page text, the text of documents you uploaded, pages captured with the clipper (including personal copies of pages you were signed in to), AI-chat conversations you saved, your notes and journals, transcripts, the AI summaries we generate for an item, and Reed's answers.

One press can send far more than one paragraph. In an article or a document, Listen reads a paragraph and then advances on its own, sending each following paragraph as the previous one finishes, until the item ends or you stop it. So a single press — even on one paragraph's own Listen button — can send the whole item. On the other surfaces one press sends one block of text, and each request is capped at the provider's 15,000-character limit; text past the cap is simply not sent.

A journal is read from what is on screen. Pressing Listen while writing a note sends the text currently in the editor, including text you have not saved.

Retention at the provider. As described in §3.4, our no-training and no-retention position for this provider rests on our agreement with them. Listen does not currently send a per-request no-retention flag, so do not treat that flag as the protection in place here. Listen never starts by itself — it always begins with a press — but see above for how far one press carries.

3.4g Share links (what a link makes public)

You can create a link to one saved item, or to a folder. Anyone holding that link can open it without an InspiMark account.

Exactly what a link publishes. For each item: its title, its description, its cover image, the original link you saved, the source (which site or platform), the site name, the author, the kind of content, and, where we have one, its length. A folder link additionally shows the folder's name, its description, how many items it holds, and the order you arranged them in, and it borrows one of those items' cover images as the folder's own preview picture. The link's own address appears in the page's preview data. That is the whole list.

Read the description before you share. The description is the field link previews display, so whatever it holds becomes the public blurb. Usually it is the source page's own preview text. Sometimes it is the item's own content: for an article saved from a full-text feed it can be the entire article; for a saved AI conversation it is the opening of the conversation; for a social post it is the start of the post. You can edit it, and a later refresh of the item may replace your edit with the source's text again.

What a link never publishes. Your notes. Your tags. Your highlights and bookmarks. Your private reason for saving something. The stored full text of an article or document, its transcript, our AI summary of it, and your Wiki. Your email address, name, or any other account identifier — a share page does not say who shared it. Pages you captured with the clipper cannot be shared at all (§2.2a), and items in your Trash drop out of a shared folder.

The Inbox folder itself cannot be shared, but an item sitting in it can be shared individually.

A link is unlisted, not secret. The address is long and random and we ask search engines not to index the page, but a forwarded link works for whoever receives it. Assume that publishing a link is publishing.

Pasting a link somewhere shares it before anyone clicks. When a link is pasted into a chat app or social network, that service fetches the page to build a preview and keeps the title, the first part of the description and the cover image. That happens whether or not a human opens the link, and the preview can outlive the link. For a video the page identifies itself as a video, so the preview reveals that a video was shared. For a folder with no description, the preview instead says how many items it holds.

Files stay private; their names may not. We never serve your uploaded files or clipped images to someone holding a share link — those requests are refused, whether the person is signed out or signed in to a different account. But a file's web address, which can contain the name you gave the file, is part of what a share link publishes: it is in the preview tags and in the public data the share page itself loads. Anyone with the link can read that address, though it will not give them the file. In a shared folder, an uploaded document's original filename is published as its description. Rename a file before sharing the folder it sits in if the name itself is sensitive.

A folder link publishes the folder as it is now. It is not a snapshot. Anything you file into that folder later is published by the same link, with no further action from you.

Taking a link down. You can revoke any share link from the share dialog, and you can turn off every share link on your account at once from your settings. Revoking takes effect immediately. You can also give a link an expiry date, or rotate it — which retires the address you already sent out and issues a new one in its place.

Deleting the shared item or folder takes its link down immediately, and deleting a folder also revokes the links on any folders nested inside it. Items inside a deleted folder keep their own links. Restoring an item from Trash brings its link back at the same address. Suspending or closing your account takes your share pages down with it.

We do not log who opens a share link. There is no view counter and no visitor record, so we cannot tell you who has seen it or how many times.

Third parties a share page does contact. Share pages carry no analytics and set no advertising cookies. They do load a web font from Google, and they show the item's cover image directly from wherever it is hosted, so Google's font service and that image's host both see the visitor's IP address. We ask browsers not to pass the link's own address along to them.

3.5 Marketing and growth of InspiMark (not sale of your library)

We may use Contact Information and coarse User Activity Information (for example, whether you completed signup or activated a feature—not the substance of private library items) to:

We do not sell your library content or build advertising profiles from what you save for third-party ads.

3.6 Legal and compliance

We use information as reasonably necessary to comply with law, legal process, or governmental requests; enforce our Terms and policies; and establish, exercise, or defend legal claims.

3.7 What we do not do with your library

3.8 Organizational / “Wiki” views

Features that organize your library by interest use content you saved to help you navigate it. They are not used to sell profiles or serve third-party ads. You can delete content or request account deletion to remove underlying data.

How Wiki pages are produced. Wiki pages are written in more than one way.

Most are written by an AI model (currently OpenAI's) as part of the same request that produces an item's summary and suggested tags after you save it. That request carries the item's text and context from the rest of your library — your tag and folder names and the titles of your existing Wiki pages — so the result is consistent with what you already have. In other words, saving one item transmits a short index of what else is in your library, though not the contents of those other items.

Pages are also written from your chat conversations. As a conversation with Reed goes on, we periodically send a summary of it and its recent turns to the model to extract Wiki pages. Because Reed's answers in a library-wide chat quote excerpts from your saved items, a page produced this way can contain material drawn from items other than the one you were looking at, and no save has to happen for it to be created.

Beyond that, you can write, edit and delete pages yourself, and an outside AI tool you have connected with a Wiki-write token can add bullets to them (Terms §4.9). Saving an item can also change a Wiki page you already have: the enrichment above may replace an existing page's key takeaways and related concepts with newly written ones.

Wiki pages are private to your account. We do not show them to other users and they cannot be shared by link, but their text is sent to our AI providers to build the embeddings that make them searchable, and a short index of your page titles is included when you start a new chat.

3.9 Deidentified and aggregated information

If we create deidentified or aggregated information that cannot reasonably be used to identify you, we may retain and use it for analytics, research, security, fraud prevention, and product improvement for any lawful purpose. We will not attempt to re-identify such data except as permitted by law (for example to test our deidentification).

4. Cookies and similar technologies

We and certain service providers use cookies, pixels, local storage, and similar technologies.

CategoryPurposeTypical requirement
NecessaryLogin/session, security, load balancing, essential preferencesRequired for the Services
FunctionalityRemember choices (e.g. UI prefs, cookie banner choice)Often necessary or low-risk
AnalyticsUnderstand traffic and product usage (marketing site and, if enabled, coarse product analytics)Consent where required (EU/UK/ePrivacy)
Targeting / advertisingInterest-based ads for our products (if we enable them)Consent / opt-out as required; off by default until configured

You can withdraw or change cookie consent via Cookie preferences on the marketing site and browser settings. Disabling some cookies may limit features.

We honor Global Privacy Control (GPC) and similar universal opt-out signals as an opt-out of sale/sharing and targeted advertising where applicable law requires it — this is a legal obligation in several US states, not a voluntary commitment. (Today the marketing site treats GPC/DNT as an analytics decline.)

5. How we share your information

We do not sell your personal information. (Under laws like the CCPA, "sale" includes exchanges for non-monetary consideration — our commitment covers those too.)

We may share information as follows:

  1. Service providers (processors). Hosting and storage (e.g. Amazon Web Services), transactional email (e.g. Amazon SES), AI/transcription APIs you invoke through the product, security and infrastructure vendors, and product or marketing analytics providers under written contracts that limit use to our instructions.
  2. Paddle, as merchant of record for paid plans. Paddle is the seller of the subscription and processes payment and tax data under its own terms and privacy notice—not as our subprocessor. We receive subscription status and limited identifiers back from Paddle.
  3. At your direction. Exports you download; OAuth integrations you connect; share links you create — a shared item or folder is viewable by anyone holding the link, without an InspiMark account, for as long as the item exists; §3.4g lists exactly which fields a link publishes and which it never does (see TOS §4.5); AI-agent access you enable — if you create a personal access token for the InspiMark MCP server, the external AI tool you give it to can search and read your library (including notes and article excerpts) and add saves and Wiki bullets on your behalf, within the scopes you chose, until you revoke the token. You choose that tool; we do not select it, cannot identify it, and cannot see what it does with what it receives, including whether it passes the text to its own AI provider (see TOS §4.9); searches you run outside your library — Discover, Trending, Smart Pulse and Monitors send your query to X, xAI, Apple, Google or Spotify depending on the feature, and Smart Pulse additionally sends the public post text currently on your screen; never your library, notes or interest profile (see §3.4b); and any future collaboration features you explicitly use.
  4. Legal and safety. As reasonably necessary to comply with law, enforce agreements, or protect rights, safety, and integrity.
  5. Business transfers. In connection with a merger, acquisition, financing, or sale of assets, subject to appropriate confidentiality and this Policy’s principles.
  6. With your consent. Where you ask us to share or otherwise consent.

A living subprocessor / vendor list will be published at launch or provided on request to privacy@inspimark.com.

6. Storage, security, and international transfers

We are based in the United States and generally process information on infrastructure in the US (and other regions we configure with our providers).

If you use InspiMark from the EEA/UK or Canada, your information may be transferred to the US. Before marketing those regions we will implement appropriate transfer mechanisms (e.g. Data Privacy Framework certification and/or Standard Contractual Clauses, and Quebec Law 25 assessments where required).

We use commercially reasonable safeguards (HTTPS, hashed passwords, access controls, optional encryption of certain secrets at rest). No method is 100% secure. Protect your password and devices. Report concerns to support@inspimark.com (or privacy@inspimark.com for privacy-related security issues).

7. Retention

DataRetention
Account & library (live systems)Until you delete items or we complete account deletion
Soft-deleted library itemsUntil permanently removed from trash / purge. Trash is not emptied automatically — deleting an item does not by itself remove it, its text, its search index or its stored file (§3.4d)
Wiki pages our AI wrote from an itemNot deleted with the item, including on permanent deletion. Delete the page yourself, or purge the account (§3.8)
Account deletion (database and associated object-storage files we control)Within 30 days of a verified deletion request (see Terms §4.6)
Embeddings / indexes / per-user transcripts and derivativesRemoved with associated content/account purge where implemented
Chat conversations (including the context stored with each question)Kept until you delete the conversation or your account is purged — deleting the item a conversation was about does not remove it (§3.4c)
Shared-cache media artifacts (transcripts/chapters/summaries of publicly accessible media)Retained after item or account deletion; removable via the takedown process
Operational DB snapshots / backupsResidual copies may remain up to 90 days (or until snapshot expiry), then age out.
Security / operator audit logs (not library body)As needed for security and compliance
User Activity logsOnly as long as reasonably needed for the purposes in §3 (shorter windows for raw logs where practical)
Paddle billing recordsHeld by Paddle per its policies and tax/legal needs
Deidentified / aggregated dataMay be retained as described in §3.9
Marketing consentsUntil withdrawn
Waitlist entries — unconfirmedDeleted automatically 14 days after the confirmation email, if the link is never clicked (§2.1a)
Waitlist entries — confirmed (email, consent record, and the network address the signup came from)Kept until you ask us to remove the entry or we retire the waitlist. Unsubscribing stops the email but does not delete the entry, and deleting an InspiMark account does not remove it either (§2.1a)
Share linksA link stays live until you delete the shared item or folder. Links do not expire on their own, and the stored link record persists with the item (§3.4g)

8. Your choices: communications, access, deletion

8.1 Opt out of non-transactional communications

Use the unsubscribe link in any marketing or product-news email — it is one click and takes effect immediately — or email support@inspimark.com or privacy@inspimark.com. We may still send transactional messages (security, billing, verification, password reset, essential service notices).

8.2 Change your information

Update profile/settings in-product when available, or email support@inspimark.com.

8.3 Export

While signed in, use in-product export tools (including JSON library export). For help or formats not yet automated, email support@inspimark.com—we aim to complete manual export help within 45 days.

8.4 Delete

Delete individual items in-product. Request account deletion by emailing privacy@inspimark.com from your account email (self-serve delete may ship later). We complete primary purge within 30 days of verification as described in the Terms. Some information may remain in backups for up to 90 days, or where law requires retention. Deidentified data may remain under §3.9.

8.5 Personalization and cookies

Disable non-essential personalization when the control is available. Manage cookies via the marketing-site preference UI and browser settings.

9. Payment processing

Paid plans are sold through Paddle as merchant of record. Card and tax details are collected by Paddle, not stored as full PANs on InspiMark servers. See Paddle’s checkout buyer terms and privacy notice, and our Terms §3.

10. Rights of US residents (CCPA and similar state laws)

If you are a resident of California, Colorado, or another US state with a comprehensive consumer privacy law, you may have rights to:

Categories we may collect (see §§2–3): identifiers (e.g. email, IP); customer records information; commercial information (subscription status); internet/network activity (User Activity Information); approximate geolocation derived from IP; audio/visual or document content you upload; inferences used only to personalize your library experience; account credentials (password is stored hashed).

Sources: you; your devices; integrations you connect; Paddle; service providers. Purposes: §3. Disclosure: service providers and parties in §5—not sold (for monetary or other valuable consideration).

Submit requests to privacy@inspimark.com. We will verify identity as reasonably necessary. Authorized agents may be required to provide proof of authority.

Working assumption at launch: CCPA “business” thresholds may not be met; we still aim for this structure. Re-test annually or at funding/growth events (see OPEN-QUESTIONS).

11. Rights of EEA and UK residents

If GDPR/UK GDPR applies, you may have rights to access, portability, rectification, erasure, restriction, objection, and withdrawal of consent (where processing is consent-based), and to lodge a complaint with a supervisory authority (or the UK ICO).

We will respond without undue delay and within one month (extendable as permitted). Contact privacy@inspimark.com. EU/UK representative: to be appointed before those markets are actively marketed; this notice will be updated with their contact details.

Lawful bases will be documented in our records of processing before launch (contract for core library; consent for non-essential cookies/marketing where required; legitimate interests for security and some product analytics, balanced against your rights).

12. AI-specific transparency

When you use AI chat or similar features, you interact with an AI system. Inputs may be sent to model-provider processors under contract to generate a response for you. Outputs can be wrong—verify important facts. See §3.4 regarding no training on your library by default.

13. Children

InspiMark is for users 16 and older. We do not knowingly collect personal information from anyone under 16. Contact us to delete if you believe we have.

14. Law enforcement

We review legal process carefully. Where permitted, we notify users of requests for their library content unless legally prohibited. Counsel will finalize our law-enforcement guidelines.

15. Changes

We may update this Privacy Notice. We will post the revised version with a new “Last modified” / effective date. For material changes we will provide a more prominent notice (in-app and/or email) as appropriate. If you disagree, stop using the Services and request deletion.

16. Contact

All privacy and DSR requests: privacy@inspimark.com Support: support@inspimark.com · Copyright / DMCA: copyright@inspimark.com · Legal: legal@inspimark.com Mail: InspiMark LLC, Attn: Nicholas Losciuto, 7483 Teasdale Ave, Saint Louis, MO 63130

EU/UK representative / Quebec privacy officer: to be appointed before those market launches.