Privacy Policy

1. Whose information this policy covers

This policy covers account holders, customers' audiences and people whose discussion content appears in a workspace.

People with a LaunchBeam account

Founders and their team members. For your account, workspace, billing and usage data we are the data controller. Everything in this policy applies to us directly, and you contact us.

People on a customer's waitlist or contact list

If you gave your email address to a product's waitlist page or received a sequence email sent through LaunchBeam, the founder who runs that product is the controller and we are their processor: we hold and send that data on their instructions and for no purpose of our own. Ask them first — but the unsubscribe link in any marketing email we deliver for them works immediately and without their involvement, and you can always write to us at [email protected] and we will route the request and help.

People whose public discussion or supplied excerpt appears in LaunchBeam

A customer may supply a discussion, or an approved discovery integration may find one. You do not need a LaunchBeam account to contact [email protected] about your content. Include the original discussion URL and the information you want removed; do not send a password. We may request proportionate evidence that you are the author or authorized to act for them. Discovery and AI processing are described below.

2. Who we are

deepfai Inc., 535 Fessler Ave, Naperville, IL 60565, United States, operating LaunchBeam. Privacy contact: [email protected]. Data protection contact: [email protected]. The service itself is governed by our Terms of Service, of which this policy forms part.

3. What we collect from account holders, and why

Account

Your email address and a password, handled by Supabase Auth — the password is hashed by Supabase and we never see or store it. Plus sign-in and account timestamps. Basis: performance of our contract with you.

Workspace and team

Workspace name and slug, who is a member, each member's role, and the email address of anyone you invite (we email them the invitation on your instruction). Basis: contract.

Launch content

Product names, descriptions and URLs, campaigns, post drafts and published posts, scheduling times, launch-directory submissions, press-kit content, testimonials, and the images and videos you upload. Basis: contract.

Platform connections

Which platform, the account handle or channel id it points at, and the credential itself — OAuth access and refresh tokens, API keys, app passwords, bot tokens. Credentials are encrypted with AES-256-GCM before they are written to the database, with the key held in the deployment environment and never in the database, and they are redacted from logs. Basis: contract.

Email sending credentials

If you configure sending, the SMTP host, port and username or the provider API key you supply, encrypted the same way, plus your from-name, from-address and reply-to, and the result of the last test send. Basis: contract.

Billing

Your Stripe customer, subscription and price identifiers, plan, billing interval and period end, and invoice metadata. We never receive or store card numbers — card details go directly to Stripe. Basis: contract, and legal obligation for tax and accounting records.

Usage and diagnostics

Counters (AI generations, campaigns created, posts published, emails sent per day), job states, and error messages returned by platforms and mail providers — which sometimes quote the content of the request that failed. Basis: contract for enforcing plan limits; legitimate interests in keeping the service working and preventing abuse.

Support correspondence

What you write to us and what we reply. Basis: legitimate interests in supporting our own product.

Referral programme

If you join LaunchBeam's own referral programme: your referral code, click counts, which accounts signed up through your link, and a commission ledger. Individual click rows record a country, a coarse device type (mobile, tablet, desktop, bot), the origin and path of the referring page with the query string stripped, and the landing path. They deliberately record no IP address, no hashed IP, and no full user-agent. Basis: contract for the programme; legitimate interests in fraud prevention.

Discussion discovery and reply drafting

A saved discussion can include its source ID and URL, community, title, body or excerpt, author username, timestamps and availability. Limited context may include the post type, comment count and whether the source is locked, archived or removed. API discovery currently searches public Reddit posts; it does not retrieve comment threads or look up author profiles. A customer can separately paste discussion or comment text. Manual input does not fetch the link or automatically verify upstream deletion.

We use deterministic rules to match discussions to the customer's product and topics, explain relevance and suggest daily actions. We keep the customer's status choices and limited workflow history. We do not use this feature to infer sensitive traits, enrich author profiles, or match Reddit users to contact lists. Discussion content is displayed within the workspace, not offered as a public mirror or bulk dataset.

Manual input requires the customer to confirm permission to supply and process the content and acknowledge the described AI processing. This acknowledgement does not establish that the original author consented and does not override platform permissions. Identified Reddit content, including manually supplied Reddit content, cannot be sent for AI drafting unless Reddit processing permission has been enabled by the operator. Customers must not disguise Reddit material as another source or include private, sensitive or unnecessary personal data.

When a customer requests an allowed reply draft, the discussion title, text or excerpt, limited context and community guidance, plus the customer's product details and brand voice, go to the configured AI providers listed in §7. A dedicated author-username field is not included in that prompt, but personal information within the supplied text may be sent; we do not promise automatic removal of every such detail. The result stays an editable draft. This workflow does not automatically post, message or vote. Opening Reddit or copying a draft does not count as a published reply.

4. Cookies and tracking

Optional analytics is off until you choose Accept. Reject is offered alongside Accept, and you can change your choice using the analytics control. We remember this choice in the lb_analytics_consent cookie for 180 days.

With permission, PostHog receives page views using a URL without its query string or fragment. Autocapture, session replay and person profiles are disabled. We do not attach your account email or account identifier to these browser events. Identifiers are held in memory rather than persistent PostHog cookies or local storage. PostHog receives requests through our own domain; this does not change its role as a third-party processor or its United States hosting.

PostHog does not run on authentication and password recovery pages, invitations, unsubscribe pages, or customer waitlist, press-kit and referral pages. Server-side product analytics is disabled unless a caller explicitly supplies consent.

After analytics consent, tracked links may set the HTTP-only lp_sid cookie for 180 days and referral links may set lp_ref for 60 days. These hold attribution identifiers and may pass a session identifier to the destination website. Without consent, links do not create attribution records or cookies.

Consent-based click records include time, link identity, referrer, browser user-agent, device category and country. These tracked-link records do not store IP addresses. Rate limiting uses short-lived IP-derived keys. This is separate from authentication session records, which can include IP addresses and user-agents for account security.

Necessary cookies include Supabase authentication cookies (sb-…) and lp_active_org, which remembers your selected workspace for up to 365 days. They support signing in and using the product.

Directory icons are rendered locally. Cloudflare provides traffic delivery and abuse protection; when configured, its Turnstile challenge processes browser and network signals on the sign-in and signup forms. Infrastructure analytics must be configured consistently with this consent policy.

5. What we process on behalf of our customers

These records belong to the founder whose workspace holds them. We process them only to deliver the features they enabled, and never for our own marketing, profiling, resale, or model training.

Waitlist entries

Email address; optional name; the referral code issued to that person and which entry referred them; a source label and any utm_source, utm_medium and utm_campaign values from the link they arrived on; a country code; and timestamps for joining, being invited, converting, and unsubscribing. Position in line is computed on demand, not stored.

Email contacts and sends

Email address; optional name; where the contact came from; whether they are still subscribed and when they opted out; and one row per scheduled or attempted send recording which sequence step it was, when it was due, whether it sent, and the error if it failed.

Affiliate and partner records

If a founder tracks launch partners: the partner's name, optional email address, the public URL of their channel, notes the founder wrote, the commission terms agreed, and conversions credited to them.

Delivery and opt-out. Mail to a customer's audience is sent through their credentials, from their domain — never from a shared LaunchBeam sending domain. Their email provider is therefore their sub-processor, chosen by them. Every marketing send carries a visible unsubscribe link and RFC 8058 List-Unsubscribe headers, so the one-click opt-out in Gmail and Yahoo works; the footer is appended to the outgoing message and a founder cannot remove it. The unsubscribe link is authenticated by a high-entropy random token alone, and the endpoint answers identically for a valid and an invalid token, so nobody can use it to test whether a given address is on somebody's list.

Outbound prospect records may contain a name, work email, company, role and source supplied by a customer. Customers attest to their permitted use and control their retention and deletion. We process these records on their instructions; scoring may send the non-email fields described in §7 to OpenAI.

6. Legal bases we rely on

• Contract — to create your account, run your workspace, publish what you ask us to publish, send what you compose, and bill you.
• Legitimate interests — security, abuse and fraud prevention, debugging, enforcing plan limits, and understanding aggregate product usage. We only rely on this where our interest does not override your rights, and you can object (§9).
• Legal obligation — keeping financial records, and responding to lawful requests from authorities.
• Consent — where required for the non-essential cookies in §4, and for any marketing email we send you about LaunchBeam itself (withdrawable at any time, without affecting service email such as billing receipts and security notices).

For subscriber data, the legal basis is the customer's to establish and document, not ours.

7. Who else processes the data

We use these sub-processors. This list includes application providers and the infrastructure that delivers the service.

Self-hosted Supabase — database, authentication and object storage

Holds essentially everything: accounts and password hashes, workspaces, products, campaigns, posts, encrypted platform credentials, waitlist entries, email contacts and send logs, analytics events, and the images and videos you upload. Hosted at Hostinger, Boston, United States (self-hosted Supabase).

Hostinger — hosting

Hosts the application, worker, Redis, database, authentication and uploaded objects in Boston, United States.

Cloudflare — network delivery and security

Processes web requests, network identifiers and security signals as our CDN and reverse proxy, and provides email routing and optional Turnstile challenges. Cloudflare Web Analytics must be disabled unless its collection is covered by the consent controls.

Stripe — payments and subscriptions

Receives what you enter in Stripe's own checkout and billing portal, including card details, which never pass through us. We receive back identifiers, plan state and invoice metadata.

Anthropic and OpenAI — draft generation

AI generation sends the relevant product, campaign or permitted discussion context and our prompt to OpenAI and/or Anthropic, according to the configured model chain. Retries and failover may send the same request to more than one provider. We do not attach platform credentials, billing records or audience contact lists to drafting prompts. For outbound prospect scoring, names, company and role can be sent; dedicated prospect email fields are excluded. Free text can still contain personal information entered by a customer. AI is used for individual inference requests, not to train or fine-tune models on discussions.

PostHog — optional product analytics

Receives consented page-view data as described in §4, hosted in its United States cloud. Session replay, automatic interaction capture and account identity binding are disabled.

Resend — LaunchBeam's own transactional email

Used for mail the platform sends about itself, and therefore receives the recipient address and the message body for each: password-reset links (a one-time credential that grants access to the account), team invitations, publish reminders, lifecycle notices and product tips you opt into. Our Resend account is deliberately not used for mail addressed to a customer's audience — though a workspace may separately choose Resend as its own sending provider below, under its own API key and its own agreement with them.

Your own email provider — audience email

The SMTP host or email API you configure receives the recipient address and message content for every waitlist welcome and sequence email your workspace sends. You choose this one.

Redis — queues and rate limiting

Short-lived job state and rate-limit counters. Email addresses are hashed before they appear in a key.

The platforms you connect

When a post publishes, its content and any attached media go to that platform under its own privacy policy and terms. That transfer is the feature.

Our OpenAI Responses requests set store: false to disable optional stored response objects. This is not zero-data-retention: provider abuse monitoring and other permitted processing can retain information separately. OpenAI and Anthropic state that API data is not used for model training by default, with opt-in or feedback exceptions. We have not verified special zero-retention arrangements for these accounts. See OpenAI's data controls, Anthropic's retention policy and Anthropic's training policy.

We will keep this list current; write to [email protected] to be notified of changes or to request a data processing agreement.

8. Who else sees the data, and who does not

We do not sell personal data, rent it, share it for cross-context behavioural advertising, or use one customer's data to benefit another. Beyond the sub-processors above, data is disclosed only:

• to a small number of LaunchBeam staff who hold platform-administrator access for support and abuse investigation. That access is granted by explicit account identifier rather than by email address, can read across workspaces, and every grant and revocation is written to an audit table;
• to professional advisers under confidentiality, where we need advice;
• to a buyer or successor in a merger, acquisition, or sale of assets — we will tell account owners before their data moves;
• to authorities where we are legally required to, and we will notify you unless the law forbids it.

9. Your rights, and how to use them

Depending on where you live you have the right to access your data, correct it, delete it, restrict or object to processing, receive it in a portable form, withdraw consent, and complain to a data protection authority. Under US state privacy laws you also have rights to know, delete, correct, and opt out of sale or sharing — we do not sell or share personal data, so there is nothing to opt out of. We do not use your data for automated decision-making that has a legal or similarly significant effect on you; the AI features write drafts and decide nothing.

To exercise a right, email [email protected] from the address on your account. We respond within one month and will say so if we need the extra two months the GDPR allows for a complex request. We may need to verify who you are, and we will not use verification data for anything else. Exercising a right never costs you service or money.

If you are on a customer's waitlist or mailing list, see §1: the founder is the controller, but the unsubscribe link works instantly and we will help either way.

Account erasure also requires reviewing uploaded objects, historical PostHog identity records and Stripe customer records. Billing records that must be retained are handled separately from optional analytics and workspace content.

10. Getting your data out, and getting it deleted

Honestly, at this stage of the product:

• Export. Waitlist entries and email contacts can be exported to CSV from the dashboard whenever you like. There is no one-click export of an entire account yet — ask us and we will produce one.
• Deletion in the product. You can delete products, contacts, images, videos, connections, and team members yourself, and deleting a record removes its dependent rows with it.
• Deleting an account or workspace. There is no self-serve delete-my-account button yet. Email [email protected] and we will delete the workspace and everything hanging off it — products, campaigns, posts, connections and their encrypted credentials, waitlist entries, contacts, send logs and analytics rows.

11. How long we keep things

• Account, workspace and launch data — for as long as the workspace exists. Deleting the workspace cascades through the database and removes its dependent records with it.
• Cookies — exactly the lifetimes in §4.
• Reddit discussion content and derived replies — expires after 48 hours from retrieval or manual addition. Re-adding the same discussion does not extend its retention. Scheduled cleanup clears source titles, text, author details, context and URLs, draft bodies and errors, and linked recommendation text, evidence and free-text feedback. Provider-observed deletion or removal can trigger cleanup earlier. Completing a reply does not extend content retention. Cleanup is bounded, so outages or backlog can delay physical removal after the content is no longer eligible for use.
• Other manually supplied discussions — the same content cleanup applies after 30 days, including answered discussions.
• Minimal discussion history, for both sources — source identifiers or hashes, ownership, workflow states, reason codes and timestamps can remain to prevent duplicates and preserve history; these records are not represented as anonymous. They remain until the workspace is deleted or a verified erasure request requires their removal.
• AI provider copies — separate from our application cleanup. Standard OpenAI abuse-monitoring data may be retained for up to 30 days; Anthropic normally deletes API inputs and outputs within 30 days. Safety, legal, contractual and feature exceptions may retain data longer. The provider policies in §7 explain those exceptions. Deleting a LaunchBeam draft does not prove deletion from an AI provider.
• Analytics and referral click rows — kept while the workspace exists. Referral visit rows are designed to be prunable at 180 days, and automated pruning of older analytics rows is not yet in place; we will not pretend otherwise.
• Unsubscribed contacts and waitlist entries — kept with an unsubscribed-at marker rather than deleted, so that a later import cannot silently re-add someone who opted out. Suppression is the reason; outright deletion is still available on request.
• Billing records — for as long as tax and accounting law requires, which is [STATUTORY RECORD-KEEPING PERIOD].
• Backups and application logs — a live-database deletion does not erase existing backups, snapshots or logs. We have not yet verified an enforced maximum age across every infrastructure copy, and do not promise that they expire within the discussion-content window. Reddit activation requires a documented, verified backup/log retention arrangement compatible with the approved use. New AI error records contain failure categories rather than model output or provider error bodies. For an erasure request, we review retained copies and apply deletion again before any restored data is returned to service.

12. Security

What we actually do:

• Platform and email credentials are encrypted with AES-256-GCM before storage, with the key held in the deployment environment and never in the database. Raw tokens are never stored and never logged.
• Every tenant table is protected by database row-level security, so one workspace cannot read another's rows even if application code gets it wrong.
• Values that identify a person are kept out of infrastructure that leaks by design: addresses are hashed before appearing in cache keys and masked in log lines, and secrets are redacted from any object we log.
• Passwords are hashed by Supabase Auth; we never receive them. Traffic is served over TLS.
• The ingest secret a customer's own application uses to post contacts to us lives in an isolated table, separate from the records it can write.

We hold no third-party security certification today — no SOC 2, no ISO 27001 — and we will not imply one. No system is perfectly secure; report anything you find to [email protected] and we will work with you.

13. International transfers

LaunchBeam is a hosted service and its sub-processors operate internationally: Stripe, Anthropic, OpenAI and Resend are United States companies, our PostHog project is on PostHog's US cloud, the database sits at Hostinger, Boston, United States (self-hosted Supabase), and the application and worker run at Hostinger, Boston, United States. If you are in the EEA, the UK, or Switzerland your data may therefore be processed outside your country. Those transfers rely on [TRANSFER MECHANISM] — such as the European Commission's Standard Contractual Clauses with the UK Addendum where it applies, or a finding of adequacy. Ask at [email protected] for the mechanism covering a specific sub-processor.

14. Children

LaunchBeam is a business tool and is not directed at children. You may not create an account if you are under 16, and we do not knowingly collect data from children under 16. If you believe a child's data has reached us, email [email protected] and we will delete it. Our customers must not use waitlist pages, contact ingest, or email sequences to collect or market to children.

15. Changes to this policy

When this policy changes, the new version appears at this URL with a new date at the top. We will email account owners before a material change takes effect, and we will never retroactively reduce your rights over data we already hold without asking you.

16. Contact

deepfai Inc., 535 Fessler Ave, Naperville, IL 60565, United States. Privacy: [email protected]. Data protection contact: [email protected]. Security: [email protected]. You also have the right to complain to your local data protection authority.