Privacy Policy

1. Who we are

SparkSocial is a social media intelligence and content management platform. In this policy, “we”, “us” and “SparkSocial” mean [REGISTERED COMPANY NAME], at —.

We are the data controller for account data and for how we run the service.

Where an agency or employer has invited you into their workspace, that organisation is the controller for the brand material and content inside it, and we act as their processor for that content. If you want content in someone else’s workspace changed or removed, ask them first — we will help them do it, but we will not override the workspace owner.

2. What we collect, and where it lives

Your identity is held by Clerk, not by us

Account sign-up, sign-in, passwords and multi-factor authentication are handled by Clerk. Your name, email address and credentials live in Clerk’s systems. SparkSocial’s own database stores Clerk’s user identifier and the role attached to it — not your password, and not a second copy of your login details.

Where the product shows a person’s name or email — on a team list, for example — it is read from Clerk at the time, not stored by us.

Brand and workspace data

• Brand profile — what you tell us about the business: category, audience, voice, offer, call to action, restricted topics and required disclaimers. Some of this may be drafted for you by reading a website address you supply, and you confirm or correct it.
• Media and documents — images, video, logos and files you upload, with the captions and usage-rights status you set against them.
• Content — drafts, scheduled posts, published posts, campaigns and calendars.
• Engagement material — comments and direct messages retrieved from accounts you have connected, and the replies drafted or sent in response.
• Trend signals — publicly available posts and topic data from supported public sources, held against your workspace where relevant to it.

Operational records

• An audit record of every action — which tool ran, who or what triggered it, what the policy engine decided and why, and what it cost. This is what lets you answer “why did this post go out?” later, and it is deliberately a projection: it records what happened, not a second copy of the content.
• Usage and cost — a credit ledger of chargeable actions, used for billing and for the usage screen.
• Technical logs — IP address, user agent, timestamps and error traces.

Where it is stored

Structured data is held in Azure Database for PostgreSQL; files and media in Azure Blob Storage; short-lived queues and caches in Azure Cache for Redis. Credentials and API keys are held in Azure Key Vault and accessed through managed identity — they are not kept in our source code or configuration files.

3. Why we use it

• To run the service — finding relevant trends, drafting content, planning the calendar, publishing to accounts you have connected, and retrieving the replies. Lawful basis: performance of a contract.
• To keep accounts secure — authentication, isolation between workspaces, abuse prevention. Legitimate interests, and legal obligation where security law applies.
• To bill you — recording chargeable actions and applying your plan’s limits. Performance of a contract.
• To keep the product working — error tracking, performance monitoring, and diagnosing faults. Legitimate interests.
• To support you — answering what you send us. Performance of a contract, or legitimate interests.
• To send you product or marketing email — only where you have opted in. Consent, which you can withdraw at any time.

We do not sell personal data, and we do not use your brand’s content to train our own models.

4. How your brand's data is kept separate

This matters most to agencies, so it is worth stating precisely.

Every piece of confidential brand material — assets, knowledge, memories, content items — is tagged with three identifiers: the organisation, the brand, and the brand’s “genome”. Every database query that touches that material must supply all three, and it is enforced in the data layer rather than by a filter in the interface. A query that tried to reach across brands is rejected rather than silently returning the wrong rows, and an automated test fails our build if any code path bypasses that layer.

The practical effect: one client’s photographs, documents and drafts cannot surface in another client’s workspace, even though both are in the same account.

People are scoped too. Beyond the account owner and administrators, a person reaches a brand only if they have been explicitly added to it.

5. Connected social accounts

When you connect an account, the platform shows you the permissions being requested and you grant them. We store the resulting access tokens — encrypted, in Key Vault-backed storage — and use them only for:

• publishing content you have approved;
• reading back performance figures for posts we published;
• retrieving comments and messages, where you have switched that on.

We do not post anything that has not cleared your workspace’s approval settings. Disconnecting an account from your workspace settings revokes our access going forward; you can also revoke it from the platform’s own security settings.

Supported platforms include LinkedIn, Instagram, Facebook, Threads, X, TikTok, YouTube, Pinterest, Reddit, Bluesky and Google Business Profile. Which are available to you depends on each platform’s own approval process and on your plan.

6. AI processing

SparkSocial uses third-party AI services to draft content, classify conversations, generate media and analyse trends. Each receives only what its task needs:

• Anthropic — the primary model provider. Receives your brand’s voice settings, the trend or brief being written about, and the draft being revised. Used for writing, classification and planning.
• OpenAI — used to turn your material into the numerical form that makes retrieval work, and as a fallback if the primary provider is unavailable.
• fal.ai — image and short video generation. Receives the prompt and the format.
• HeyGen — avatar video, where you have enabled it and recorded consent.
• ElevenLabs — voice-over and dubbing.
• AssemblyAI — captions and transcription from media you supply.

Everything these produce is a draft. It is shown to you for review and does not publish without clearing your workspace’s approval settings.

Avatar and voice cloning are off by default and require a recorded consent entry naming who the likeness belongs to and who granted permission. That record is kept append-only: revoking consent writes a new fact rather than deleting the history of what was once permitted.

[CONFIRM AND STATE EACH PROVIDER'S TRAINING POSITION] — each provider’s terms govern whether data sent to them may be used for their own model training. We have selected providers on the basis that customer content is not used for training; confirm the current commercial terms of each and state them here before you publish.

7. The WhatsApp capture loop

If you use the capture loop, SparkSocial sends shot lists to a phone number you provide and receives the clips and photographs you send back, through the WhatsApp Business Platform operated by Meta. Meta’s own terms and privacy policy apply to the transport of those messages.

The phone number is used only for that loop. In our internal audit records the recipient number is redacted before the record is written, so the operational history of who was messaged does not accumulate readable phone numbers.

8. Sub-processors

These are the services that process data on our behalf. Each is bound by a data processing agreement.

Infrastructure

• Microsoft Azure — hosting, database, file storage, cache, secret management.
• Trigger.dev — scheduling and long-running background work.
• Mux — video hosting and delivery.

Identity and communication

• Clerk — accounts, authentication, organisation membership.
• Meta (WhatsApp Business Platform) — capture-loop messaging.

AI and media

• Anthropic, OpenAI — language models and embeddings.
• fal.ai, HeyGen, ElevenLabs, AssemblyAI — image, video, voice and transcription.

Publishing and links

• The platforms you connect — LinkedIn, Meta, X, TikTok, Google, Pinterest, Reddit, Bluesky.
• Ayrshare — a publishing aggregator, used only for platforms we do not reach directly.
• Dub — short links and click measurement, where you use tracked links.
• Canva — where you connect a Canva account for design assets.

Diagnostics

• Sentry — error reporting.
• PostHog — product analytics, on its EU instance.
• Langfuse — AI call tracing. See the next section for what it does not receive.

9. Diagnostics, and what we deliberately do not send

Monitoring how the AI behaves is necessary to keep it working. Sending your material to a monitoring vendor to do it is not.

Our tracing pipeline runs on an allow-list: only named, non-sensitive fields — identifiers, format names, counts, decisions — are sent to Langfuse. Free text is not on that list. Post copy that has not been published, asset captions, phone numbers and brand material do not leave for tracing purposes.

The direction of failure is deliberate. If a new field is added to the product, it is invisible in our tracing until somebody explicitly adds it — which is an inconvenience for us rather than a disclosure for you.

10. How long we keep it

Workspace content is kept until you delete it or close the account. There is no automatic expiry unless you set one.

Each workspace can set its own retention period, after which content is removed automatically. Leaving it unset means “keep indefinitely”, and that is the default — a product that quietly started deleting customer material would be worse than one that keeps it until asked.

• Workspace content — until deleted, or per your retention setting. On account closure, deleted or anonymised within [PERIOD].
• Audit and cost records — retained for accountability and billing for [PERIOD].
• Technical logs — [PERIOD].
• Backups — deleted content persists in backups for [PERIOD] before being overwritten.

11. Your rights

Subject to where you live, you may:

• ask what personal data we hold about you, and receive a copy;
• have inaccurate data corrected;
• ask for data to be deleted;
• object to, or ask us to restrict, certain processing;
• receive your data in a portable form — the product also has a brand export you can run yourself;
• withdraw consent where we relied on it;
• complain to a supervisory authority. In Nigeria that is the Nigeria Data Protection Commission; in the EU or UK, your national authority.

Email — and we will respond within the period the applicable law requires. If your data sits inside an organisation’s workspace, we will pass the request to that organisation and support them in answering it.

12. Security

• Data encrypted in transit and at rest.
• Secrets in a managed vault, reached by managed identity — never committed to source control.
• Brand isolation enforced in the data layer and covered by a test that fails the build if bypassed (section 4).
• Role-based access, with approval rules a workspace can require before sensitive actions.
• An audit record of every action taken in your workspace, by a person or by the agent.
• Access limited to the staff who need it to operate and support the service.

No system is perfectly secure and we do not claim otherwise. If a breach affects your personal data, we will notify you and the relevant authority within the time the law requires.

13. Where data is processed

Our infrastructure runs on Microsoft Azure in [PRIMARY REGION]. Several of the processors in section 8 operate in other countries, including the United States.

Where personal data is transferred internationally we rely on the safeguards available under the applicable law — standard contractual clauses, an adequacy decision, or the transfer conditions in the NDPA.

Enterprise workspaces can record a data residency preference. Where we cannot meet a stated preference for a given feature, we will say so rather than silently process elsewhere.

14. Cookies

This marketing site sets no analytics or advertising cookies. The product uses:

• Essential — session and security cookies set by Clerk, and a cookie remembering which brand workspace you are viewing. These cannot be switched off without breaking sign-in.
• Analytics — PostHog, to understand which features are used. Set only where you have consented, if consent is required where you are.

You can clear or block cookies in your browser, but you will not stay signed in.

15. Still to be decided

These are open business decisions, not descriptions of practice. They are listed here so nothing on this page is mistaken for settled policy:

• The registered company name and, if one is appointed, a data protection contact.
• The four retention periods in section 10.
• The primary Azure region in section 13.
• Confirmation of each AI provider’s training terms (section 6).
• The minimum age for an account — SparkSocial is a business product and is not directed at children, and we do not knowingly collect data from anyone under [AGE].

16. Changes

We will update this page when our practices change and revise the date at the top. Where a change materially affects you — a new category of processing, or a new sub-processor handling your content — we will tell you by email or in the product before it takes effect.

17. Contact us

• Email: —
• Phone: —
• Post: —