Your context stays yours.

Summary

Solum is a personal context vault. You put things into it — notes, facts, preferences, saved AI conversations, files — and you connect AI assistants that can read from it through the Model Context Protocol (MCP).

Three things are worth knowing before you read the detail:

1. We can read most of what you store. Solum encrypts your content at rest, but for the majority of your vault we hold the encryption keys. That is what makes the product work — an AI assistant asks our server for your context and our server has to be able to read it in order to answer. Solum is not end-to-end encrypted and is not a zero-knowledge service. A narrow category of content — entries you have moved into our client-side encrypted tier by enrolling a passkey and running the re-encryption step — is sealed with a key we do not hold. Everything else, including content merely labelled high sensitivity, is readable by us. §5 explains this precisely, because the distinction is easy to get wrong and matters a great deal.
2. We do not sell your data and we do not train AI models on it. Not our models, and — by contract — not our vendors' models either.
3. When you connect an AI assistant, content leaves us. Once your context is delivered to ChatGPT, Claude, or any other assistant you connect, that vendor's privacy policy governs what happens to it. We cannot pull it back.

This summary is not a substitute for the sections below.

---

1. Who we are

Solum is operated by Nathan Ajmera, an individual doing business as Solum, located in Los Angeles, California, United States.

For the purposes of the EU General Data Protection Regulation (GDPR) and the UK GDPR, Nathan Ajmera is the data controller for personal data processed through Solum.

Contact for all privacy matters: solumverif@gmail.com

Data Protection Officer: We have not appointed a Data Protection Officer. We are not required to under GDPR Article 37, because our core activities do not consist of large-scale systematic monitoring or large-scale processing of special categories of data.

---

2. The personal data we collect

2.1 Account and identity data

2.2 Vault content — the substance of the product

This is the category that matters most, and it is entirely under your control in the sense that you decide what goes in.

• Facts, preferences, values, relationships, skills, and ongoing context you enter directly or approve from a suggestion.
• AI conversations captured from assistants you connect — the full text of your messages and the assistant's replies, saved verbatim, together with a generated title and summary.
• Your Talk to Solum conversations. Conversations you have with Solum itself are captured into your vault on the same terms as conversations with any assistant you connect. They become ordinary vault entries: searchable, editable, and deletable by you.
• Notes and snippets you write or paste in.
• Files you upload, including documents and PDFs, and the text extracted from them.
• Projects, spaces, and domains you use to organize the above.
• Sensitivity labels you assign (low, medium, high) and any redaction masks you apply.
• Vector embeddings — numeric representations of your content used to make search work.

Vault content can contain anything you put in it, including data that is "special category" data under GDPR Article 9 (health, religious or philosophical beliefs, political opinions, trade union membership, sex life or sexual orientation, biometric or genetic data) and data about other people — colleagues, family, clients. See §4.4 and §5 for how we treat that.

2.3 Connected-account data (optional connectors)

If — and only if — you connect a Google or Microsoft account, we process:

• OAuth access and refresh tokens for the accounts you connect.
• A metadata index of the items in those accounts. For each item this is: an identifier, a title (email subject, file name, event name, or space name), a snippet of up to approximately 200 characters taken from the beginning of the body, the sender or owner, timestamps, and provider metadata.
• Sync bookkeeping — provider change cursors, subscription state, and sync error records.

We do not store the full body of any connected item. The index table has no column capable of holding one — it holds a title and a short snippet, and nothing else. Full message bodies, complete file contents, and attachments are never written to our database.

When a connected AI assistant asks about an item found through the index, we return that item's metadata together with a routing instruction telling the assistant to retrieve the body through its own connection to Google or Microsoft. The body does not travel through us.

2.4 Connection, activity, and audit data

• Connected AI records — which assistants you have connected, when, what permissions you granted, and the OAuth client registrations behind them.
• An audit log — an append-only record of every change to your context: who or what made it (you, an MCP client, the browser extension, or a system job), the change type, a diff of what changed, cryptographic hashes of the before and after state, and a one-line human-readable summary.
• Retrieval and delivery records — which entries were served to which assistant, and when.
• Usage counters — how many operations, chat messages, and AI-assisted actions you have used against your plan's limits, and the associated API cost.

2.5 Technical data

• Server logs generated by our hosting and database providers, including IP address, user agent, timestamps, and request paths, retained by those providers under their own policies.
• Crash reports — when Solum throws an error, our error-monitoring provider (Sentry, §6.2) receives the error message, the stack trace, the runtime it happened in, the request method, and the URL path with query strings and record identifiers removed. We never attach your name or your email address, and the only identifier that can accompany a crash report is your account's internal ID. Request bodies, cookies, headers, and console output are stripped before the report leaves your browser or our server, and we have switched off that provider's storage of IP addresses. Crash reports are deleted within 90 days at the outside; on our standard plan the window is 30 days, and it is longer only while we are on a higher-tier plan or trial.
• Session cookies — see §11.

2.6 What we deliberately do not collect

• We do not use third-party analytics, advertising, or tracking services. There is no Google Analytics, no advertising pixel, no session-replay tool, and no cross-site tracker on Solum. We do use one crash-reporting service — Sentry — which receives a report when Solum breaks. It is not analytics: it cannot see what you do when nothing goes wrong. Sentry's session-replay and performance-tracing features are disabled in our code rather than merely left unused, so no recording of your screen or your clicks is ever captured. See §2.5 and §6.2.
• We do not collect precise geolocation.
• We do not collect payment card data. Solum is free today and no payment processor is connected to the Service. If we introduce paid plans, payment will be handled by a PCI-compliant processor, card data will never reach our servers, and we will name that processor here before the first charge.

---

3. Where your data comes from

---

4. Why we process your data, and our legal bases

Under GDPR and UK GDPR we must tell you the purpose and the lawful basis for each kind of processing. Here they are, paired.

Providing your data is not a statutory requirement. It is a contractual necessity: you cannot use Solum without an account, and the vault is empty until you fill it. You are free not to use the connectors at all; the service works without them.

4.1 We do not train AI models on your content

We do not use your content — vault entries, saved conversations, uploaded files, or anything obtained through a connector — to train, fine-tune, or improve any artificial intelligence or machine learning model.

We use third-party AI APIs to deliver features to you (see §6). We use those providers' commercial API terms, under which content submitted through the API is not used to train their models. We do not opt in to any data-sharing or model-improvement program offered by any AI provider we use.

A precise clarification, because the distinction matters: we generate vector embeddings of your content so that search works. An embedding is a numeric representation of one specific entry in your vault, stored against your account, used only to retrieve your own content for you. It is not a model, it is not shared, and it is not used for anyone else. We consider this part of delivering the service to you, not model training — but we would rather describe it plainly than let you discover it later.

4.2 Automated decision-making

Solum does not make decisions about you that produce legal effects or similarly significantly affect you within the meaning of GDPR Article 22 or UK GDPR Articles 22A–22D. Our AI features suggest, summarize, classify, and retrieve; they do not decide anything about your rights, your access to services, your employment, your credit, or your entitlements.

We do use automated processing to suggest facts from your conversations, to classify content into projects and sensitivity levels, and to rank search results. You can review, edit, or delete any of it.

4.3 Profiling

We build a profile of you in the ordinary-language sense — that is the entire product. We do not profile you for advertising, scoring, or any decision made about you by us or anyone else.

4.4 Other people's data in your vault

When you save a conversation, connect a mailbox, or upload a file, that content will often contain personal data about people who are not Solum users and who have not agreed to anything with us. You are responsible for having a lawful basis to put that data into Solum, and the Terms of Use say so. We process it as a processor acting on your instructions to the extent that framing applies, and we apply the same protections we apply to your own data. If a third party contacts us about their data in someone's vault, we will generally direct them to the account holder, and we will act on our own where the law requires it.

---

5. How we protect your data — stated precisely

We would rather tell you exactly what our security posture is than use a marketing word that overstates it. Solum is not end-to-end encrypted. Solum is not zero-knowledge. We hold keys that let us read most of your vault.

5.1 Two storage tiers, two different guarantees

Solum stores content in one of two tiers. The tier, not the sensitivity label, is what determines whether we can read something. These are related but they are not the same thing, and conflating them would give you a false sense of security.

Tier 1 — the default, and where nearly everything lives. We can read it.

Tier 1 content is encrypted at rest with AES-256-GCM using a key that we hold. Our servers decrypt it in order to search it, rank it, and deliver it to assistants you have connected. This is structural, not an oversight: MCP works by an assistant asking our server for your context while you are not present, and a server that cannot read the content cannot answer.

Content labelled "high sensitivity" is still Tier 1 unless you have taken the extra step described below. A high-sensitivity label changes how the content is encrypted at rest — it is sealed with our server key rather than stored as plain text — but we still hold that key and can still read it, and it is still eligible to be served to AI assistants over MCP.

What this means concretely: our infrastructure can read your Tier 1 content. A breach compromising both our database and our encryption key would expose it. A valid legal order could reach it.

Tier 2 — content sealed with your key. We cannot read it.

Tier 2 content is encrypted in your browser with a master key derived from your passkey (WebAuthn) or from a passphrase you set (Argon2id). That key never reaches our servers. We hold ciphertext and nothing else. Tier 2 content is never served over MCP to any AI assistant, is never sent to any AI provider by us, and is never decrypted on our servers. It surfaces only in your own unlocked session, where your device decrypts it and you decide, per prompt, what to use.

How content actually gets into Tier 2 — read this carefully.

Tier 2 is not automatic. To put content there you must:

1. enrol a passkey, or set a passphrase, on the security page — this creates the master key; then
2. label the item high sensitivity; then
3. run the re-encryption step on that same security page, which pulls the item's plaintext down to your browser, encrypts it under your master key, and sends back ciphertext that replaces our readable copy.

Until you complete step 3 for a given item, that item is Tier 1 and we can read it — regardless of its sensitivity label. There is no automatic path that seals a newly created high-sensitivity item into Tier 2; the re-encryption step is the only route, and you must run it again for content you add later.

The trade-off, once content is in Tier 2: if you lose your passkey, forget your passphrase, and lose your recovery codes, that content is permanently unrecoverable. We hold no escrow copy of your master key and no way to derive it. Your recovery codes are stored only as a lookup hash that cannot reconstruct the key. We cannot reset it for you under any circumstances, including a legal order. That is the point of the design.

Two further caveats, because they are easy to miss.

Timing. Anything that existed as Tier 1 before you re-encrypted it was readable by us during that period, and will already have been indexed for search — meaning its text was processed by our embedding provider. Moving it to Tier 2 protects it going forward. It cannot un-send anything already sent.

Automated labelling. Solum's AI-powered fact extraction proposes a sensitivity level for the facts it suggests, including "high." Where a label was proposed that way, the content was necessarily processed by an AI provider in order to be assessed. Labels you set yourself do not involve that step.

5.2 Connector tokens — an honest exception

Your Google and Microsoft OAuth tokens are stored twice, under two different keys:

• once under your passkey-derived key, for use by your own unlocked browser; and
• once under a key we hold, so that background sync can refresh your metadata index while you are offline.

This means we can read your connector tokens, and a breach of our database together with our server key would yield usable tokens for connected accounts. We use the server-readable copy for one purpose only: walking the provider's change feed to write metadata rows into your index. We do not use it to fetch or store item bodies.

Background sync is not separately optional — connecting an account enables it. There is no toggle that connects an account without it. If you do not want it, the remedy is to disconnect the account, which removes the stored tokens and the index built from them (§7.4), or not to connect it in the first place.

5.3 Everything else we do

• In transit: TLS for all connections to Solum.
• At rest: AES-256-GCM for vault content and tokens; our database provider encrypts the underlying storage as well.
• Access control: row-level security on every table that holds your data, scoped to the owning account. Team administrators have no read policy over member vaults.
• Authentication: WebAuthn passkeys, Argon2id for passphrase-derived keys, HttpOnly cookies, CSRF protection, PKCE on every OAuth flow.
• Audit: an append-only log of every change to your context and every delivery to an assistant, visible to you and exportable.
• Human access: Solum is operated by one person. We do not routinely read your content. We access it only where you ask us to for support, where we must investigate a specific security incident or suspected abuse, or where the law requires it.

5.4 What we cannot promise

No service can guarantee security. We do not claim to be unhackable, bank-grade, military-grade, or fully secure, and you should be skeptical of any service that does. If a breach affects your personal data, we will notify you and the relevant supervisory authority as required by law — under GDPR, without undue delay and where feasible within 72 hours of becoming aware.

---

6. Who we share your data with

We do not sell your personal information. We do not share it for cross-context behavioral advertising. We have never done either, and we do not have a business model that would benefit from it.

We share data in exactly four situations.

6.1 AI assistants you connect — the important one

When you connect an AI assistant through MCP, or use a Solum feature that calls an AI model, your Tier 1 content can be sent to that assistant or provider so it can answer you.

Once content reaches an AI assistant you connected, that vendor's privacy policy governs it, not ours. We cannot retrieve it, delete it from their systems, or control their retention. You choose which assistants to connect and you can revoke any of them at any time from your dashboard; revoking stops future deliveries but does not undo past ones.

Tier 2 content is never delivered this way.

6.2 Our service providers (subprocessors)

Each of these processes data on our instructions under a data processing agreement.

We will update this table before adding a new subprocessor that handles vault content, and we will tell you about it under §14.

6.3 Your organization, if you joined through one

If you use Solum through an employer's team plan:

• Your administrator can send content to your company account (a "broadcast"). Broadcast entries are marked as company reference material and are kept separate from your personal facts.
• Your administrator can see aggregate counts only — how many members are active, how many entries exist in total. Administrators cannot read your entries. This is enforced in the database, not just in the interface.
• If your administrator revokes your seat, your company account is deleted in full after a 7-day grace period, along with everything in it. Your separate personal account, if you have one, is untouched.

6.4 Legal and structural

• When the law requires it — in response to a valid legal process. We will notify you unless we are legally prohibited from doing so.
• To protect rights and safety — where we reasonably believe disclosure is necessary to prevent serious harm or investigate fraud or abuse.
• In a business transfer — if Solum is acquired or its assets are transferred, your data may be part of it. We will notify you before that happens and give you the opportunity to export and delete your data first. For data obtained through Google Workspace APIs, we will obtain your explicit consent, as Google's Limited Use requirements demand.

---

7. Google user data

Solum's use of information received from Google APIs adheres to the Google API Services User Data Policy, including the Limited Use requirements.

7.1 The scopes we request and why

We request read-only access in every case. Solum never writes to, modifies, or deletes anything in your Google account.

7.2 What we do with Google data

• We store a metadata index only: identifier, title, sender, timestamps, and a snippet of up to approximately 200 characters.
• We do not store message bodies, full file contents, or attachments. When an assistant needs the full content of an indexed item, we return the item's metadata and direct the assistant to retrieve it through its own connection to Google. The body is not fetched by our servers and is not stored by us.
• We use the data only to provide the search, retrieval, and assistant-context features that are visible and prominent in Solum's interface.

7.3 What we never do with Google data

• We do not retain, use, or transfer information obtained through Google Workspace APIs to develop, improve, or train generalized or non-personalized artificial intelligence or machine learning models.
• We do not transfer or sell Google user data to advertising platforms, data brokers, or information resellers.
• We do not use it to serve advertisements.
• We do not use it to determine credit-worthiness or for lending purposes.
• We do not allow humans to read your Google data, except: where you have given affirmative agreement for us to view specific items (for example, when you ask for support on a specific problem); where it is necessary for security purposes, such as investigating abuse; or where required to comply with applicable law.

7.4 Revoking access

You can revoke Solum's access to your Google account at any time at myaccount.google.com/permissions. Doing so immediately stops us from reading anything further from your account.

You can also disconnect a connector from your Solum Connections page at any time, and this is the more complete route.

When you disconnect, we always delete the stored OAuth tokens for that connector, and — where the provider supports it and no other connector is still using the same grant — we revoke the grant at the provider on your behalf. Google supports this and we use it. Microsoft does not offer delegated-consent revocation to applications, so for Microsoft we delete our tokens and send you to your Microsoft account page to complete the revocation yourself. Where we cannot revoke for you, we say so and give you the link rather than implying it is done.

You choose what happens to the metadata index we already built. Disconnecting offers two options:

• Delete it — we purge every indexed item for that connector, along with its sync state and activity history. Nothing about that connector's contents remains.
• Keep it — the index stays searchable in Solum, but nothing further is read from your account. Choose this if you want past results to keep working after you disconnect.

Revoking at the provider without disconnecting in Solum stops all further reading, but the tokens and the metadata index already built remain until you disconnect or delete your account.

---

8. Microsoft user data

If you connect a Microsoft account, we request the delegated Microsoft Graph permissions Files.Read.All and User.Read, plus standard sign-in scopes. Both are read-only.

We treat Microsoft data on exactly the same terms as Google data: a metadata index only, no file bodies stored, no model training, no sale, no advertising, and human access only in the narrow circumstances described in §7.3. You can disconnect the connector from your Solum Connections page, which deletes the stored tokens and lets you purge or keep the index. Microsoft does not let an application revoke a delegated consent on your behalf, so you must also revoke it yourself at myaccount.microsoft.com — Solum links you there at the point of disconnect. §7.4 otherwise applies equally here.

---

9. How long we keep your data

9.1 Deleting your account

You can delete your account from Settings. Deleting your account removes your authentication record, which cascades to every table holding your data — vault entries, facts, preferences, values, files, projects, spaces, captures, connector metadata indexes, connector OAuth tokens, Talk to Solum history, connected-AI records, audit log, and usage counters. Your login identity is deleted with it, so your email address does not survive the deletion.

Deletion is immediate and cannot be undone. There is no grace period and no restore. Export anything you want to keep first — see §12.1.

Backups held by our infrastructure providers may retain deleted data for a short period before rolling off.

---

10. International data transfers

Solum is operated from the United States and all processing happens in the United States.

If you are in the EU or UK: you are providing your data directly to a US controller. Under the European Data Protection Board's guidance on the interaction between Article 3 and Chapter V, your own direct disclosure to us is not a "restricted transfer," because there is no EU-established exporter. GDPR still applies to us in full under Article 3(2).

For onward flows to our service providers, where a transfer mechanism is required we rely on the European Commission's Standard Contractual Clauses (Implementing Decision (EU) 2021/914) and, for the UK, the International Data Transfer Agreement or the UK Addendum. Our providers — Supabase, Vercel, Anthropic, OpenAI, Google, Microsoft, and Sentry — each make these available in their standard data processing terms.

We do not claim certification under the EU–US Data Privacy Framework. We are not certified, and we will not say otherwise.

You may request a copy of the safeguards in place for any transfer by writing to solumverif@gmail.com.

The United States does not have a general data protection law equivalent to GDPR, and US authorities may in some circumstances access data held by US providers. We disclose this so your consent is informed.

---

11. Cookies, Do Not Track, and Global Privacy Control

Cookies. Solum uses only the cookies it needs to work: session and authentication cookies set by our authentication provider, short-lived HttpOnly cookies carrying OAuth state during a connector handshake, and a small number of preference cookies. We do not use advertising cookies, analytics cookies, or any third-party tracking cookie. Because all of our cookies are strictly necessary to deliver a service you requested, we do not display a consent banner; if we ever add a non-essential cookie, we will ask first.

Do Not Track. Some browsers send a "Do Not Track" signal. There is no industry consensus on what a service must do in response. Solum does not track you across other websites or over time in the first place, so there is nothing for the signal to switch off. We do not respond to DNT signals differently from ordinary requests.

Global Privacy Control. We honor GPC signals. Because we do not sell or share personal information for cross-context behavioral advertising, a GPC signal does not change how we handle your data — there is nothing to opt out of.

---

12. Your rights and choices

12.1 Rights available to everyone, regardless of where you live

We extend these to all users, not only those in jurisdictions that require them.

• Access — get a copy of the personal data we hold about you.
• Correction — fix anything inaccurate. Most of it you can edit directly.
• Deletion — delete individual entries, disconnect connectors, or delete your entire account.
• Portability — export your entire vault and your audit log as JSON from Settings, at any time, in a machine-readable format. Content you have sealed into Tier 2 is listed in the export but stays encrypted, because we cannot decrypt it; unlock it in the app to read it.
• Restriction and objection — ask us to limit or stop a particular processing activity.
• Withdraw consent — revoke any connected AI at any time from your dashboard, disconnect a connected Google or Microsoft account from your Connections page, and revoke the grant at the provider itself. Withdrawal does not affect the lawfulness of processing before you withdrew.
• Non-retaliation — we will not degrade your service or charge you differently for exercising any of these.

Most of this is self-service in the app, which is the fastest route. For anything else, write to solumverif@gmail.com.

How we verify a request: we will ask you to make the request from the email address on the account, or to demonstrate control of the account by signing in. For a deletion request we may ask for a second confirmation, because it is irreversible. We do not ask for government ID.

How fast we respond: within 30 days. If a request is complex we may extend by a further 60 days and will tell you why within the first 30. We do not charge for this unless a request is manifestly unfounded or excessive.

Authorized agents: you may use an authorized agent. We will ask for written proof of authorization and may still verify with you directly.

12.2 If you are in the EU or UK

You have the rights above under GDPR Articles 15–22 and the equivalent UK GDPR provisions.

Complain to us first. If you are unhappy with how we handle your data, email us at solumverif@gmail.com with "Data protection complaint" in the subject line. This is our electronic complaints route and it is monitored. We will acknowledge your complaint within 30 days and respond without undue delay.

Complain to a regulator. You also have the right to complain to a supervisory authority — in the UK, the UK data protection regulator (currently the Information Commissioner's Office, being reconstituted as the Information Commission), at ico.org.uk/make-a-complaint; in the EU, your national data protection authority. You do not have to complain to us first.

12.3 If you are in the United States

We are a very small business and do not meet the applicability thresholds of the California Consumer Privacy Act or of any other state comprehensive privacy law. We nonetheless extend the substance of those rights to every US user, as set out in §12.1.

Specifically and affirmatively:

• We do not sell your personal information, as "sell" is defined in the CCPA or any state privacy law, and we never have.
• We do not share your personal information for cross-context behavioral advertising, and we never have.
• We do not use or disclose sensitive personal information for any purpose other than providing the service you asked for.
• We do not process personal information for automated decision-making that produces legal or similarly significant effects.

California residents additionally have rights under California's "Shine the Light" law (Civil Code § 1798.83) to request information about disclosures to third parties for direct marketing. We make no such disclosures, so there is nothing to report.

12.4 If you are in Canada

You have rights of access and correction under PIPEDA, and may complain to the Office of the Privacy Commissioner of Canada.

---

13. Children's privacy

Solum is not for anyone under 18. You must be at least 18 years old to create an account. We do not knowingly collect personal data from children. If we learn that we have, we will delete the account and its data promptly. If you believe a child has created an account, write to solumverif@gmail.com and we will act.

---

14. Changes to this policy

We will change this policy over time. How we notify you depends on what changed.

For clarifications, corrections, and changes that are neutral or better for you: we will post the updated policy at usesolum.co/privacy with a new "Last updated" date and version number. We keep every prior version and will send you any of them on request.

For any change that expands how we use your data — a new purpose, a new category of recipient, a new subprocessor handling vault content, or anything that makes our practices more permissive:

1. We will email you at the address on your account at least 30 days before the change takes effect, and show an in-app notice.
2. The new use will apply only to data collected after the change takes effect, not retroactively to data you have already given us.
3. Where the change is material and requires your consent, we will ask for it affirmatively. We will not treat your silence or your continued use of the service as agreement to a materially more permissive data practice.

We are committing to this because the alternative — quietly rewriting the policy and calling continued use consent — is the practice regulators have singled out as unfair and deceptive, and because you gave us your data on the basis of what this document says today.

---

15. Contact us

Privacy and data protection: solumverif@gmail.com Controller: Nathan Ajmera, doing business as Solum, Los Angeles, California, United States

We aim to respond to every privacy enquiry within 30 days.

---

Solum · usesolum.co · Version 2.2 · Effective August 6, 2026