The answers your compliance reviewer will ask for.
Written to be forwarded. Every answer here is one you can send straight to the person at your firm who has to sign off, or to your client's counsel at NDA stage.
Frankfurt, Germany (fra1)
Frankfurt, Germany (eu-central-1)
None
Who is the data controller and who is the processor?
You are the controller of the personal data you put into your workspace - your contacts, your counterparties, your notes. DealLaunch AI Ltd is the processor, acting only on your instructions. Our Data Processing Agreement records this and is available to sign before you upload anything.
Where is the data processed?
Stored in the European Union. Application server functions run in Frankfurt, Germany (fra1), and the database, authentication and document storage sit in Frankfurt, Germany (eu-central-1) - the same city, so the stored copy of your deal data stays in the EU. Some of it is sent to providers in the United States, under the standard contractual clauses in their own terms. Anthropic receives the project context for the drafts, summaries and research you ask for, and the rows of a contact spreadsheet you import, which it tidies. RocketReach receives a firm name and a person's name or role for a contact search, and Anymail Finder a person's name and the company's domain or name when you look up one named person's address. No AI provider is asked to find or guess an email address. On projects where automatic reply status is switched on, TypeSafe AI receives each email from the project's contacts found in your mailbox, and each reply to your outreach from a colleague of theirs at the same firm, with the names of its attachments, checked every ten minutes, and every five minutes while Outreach is open. None of them has access to your database or your document store.
Is workspace data separated between customers?
Yes. Every table carries the owning workspace and is protected by Postgres row-level security, so a query authenticated as one firm cannot read another firm's rows. The one exception is a shared table of company email formats - a firm's domain and address shape, learned from paid address lookups on Anymail Finder or RocketReach - which holds no names, addresses or workspace data and is readable only by our server. Uploaded documents live in a private storage bucket reachable only through short-lived signed links.
Who at DealLaunch can see our deal data?
Nobody routinely. There is no customer-support console that reads your workspace. Direct database access is used only for maintenance and incident response, and we will tell you if your workspace was involved in an incident.
How is outreach sent, and does it touch your servers?
Your outreach has no mail sub-processor. It is sent directly from your own mailbox - through Microsoft Graph, the Gmail API, or your firm's own SMTP server - using credentials encrypted with AES-256-GCM that are decrypted only on our server. Nothing you send to a counterparty is relayed through DealLaunch infrastructure, which means your firm's existing journaling and archiving still capture every message. Each outreach email carries a 1x1 image from DealLaunch, and when the recipient's mail app loads it we record the time against that contact and that email - never their IP address or browser. Our own transactional email - a team invite, telling somebody their access was approved, or a document or reply alert you have not switched off - is sent through Resend, a listed sub-processor. It carries no deal data, except that a document alert names the counterparty, their firm and the document type, and a reply alert names the counterparty and their firm. Replies are a different path: when replies are checked - when you open Outreach, and every 10 minutes on our server for a mailbox connected through Microsoft or with an app password - our server reads email from your project contacts in your mailbox and saves its subject and opening text on the contact's record, and on a project where automatic reply status is switched on it also sends that email's subject and text, with the names of its attachments (never the files), to TypeSafe AI, a listed sub-processor, to set the contact's status. It treats a reply to your outreach from someone else at the contact's firm, written from that firm's own company domain, the same way, as that firm's reply. On those projects our server also reads, itself and with no AI, e-signature notices about your project contacts and the mail you sent them - for its recipients, subject, attachment names and DealLaunch links - to record a signed NDA or a CIM.
Is our data used to train AI models?
No. We do not train models on your data. AI drafting, summarising and research run when you ask; importing a contact spreadsheet sends its rows to Anthropic to tidy them; and on a project where automatic reply status is switched on, TypeSafe AI's Jev model reads each email from that project's contacts, and each reply to your outreach from a colleague of theirs at the same firm, with the names of its attachments, to set their status. Anthropic's business terms prohibit training on what we send. Anymail Finder, asked for one named person's address, receives only that name and the company, and its terms do not restrict its own use of search data. TypeSafe AI's terms prohibit training on it without our consent, which we have not given, but let TypeSafe use telemetry from it - logs, classifications, metrics and learnings - without restriction, including to improve its products.
Can we get our data out?
Yes, whenever your account is active, without asking us. Settings -> Security exports what your login can see as a JSON file: deals, contacts outside the trash, document records, drafts, templates and the activity record. It holds only those records, so uploaded files and comments are not in it - a file opens in the product while it is in your workspace, and we return or erase files if you ask at hello@deallaunch.ai. If your access has ended, write to the same address. There is no exit fee and no notice period.
What happens if DealLaunch stops trading?
Your records are exportable on demand today, which is the only answer that actually protects you. We recommend taking a periodic export and keeping it with your own records, along with your own copies of uploaded files, exactly as you would with any vendor.
Is there an audit trail?
Every material change to a deal, contact, document or report writes an append-only activity record with a timestamp and the person who made it, and it is included in your export. A separately hardened, tamper-evident audit log with a dedicated actor field is in progress - see the honest list below.
Do you hold SOC 2 or ISO 27001?
No, and we will not imply otherwise. We are a young company. Our infrastructure providers hold those certifications for the layers they operate; we do not hold one for ours. If a certification is a hard requirement for your firm, tell us before you trial the product.
Is DealLaunch a regulated firm?
No. DealLaunch AI is workflow software. It does not provide investment advice, investor matching, brokerage, placement agency or any regulated financial service, and it must not be relied on as any of those.
Sub-processors
The complete list, published rather than held back for procurement. The full table, with the notification commitment, is on the sub-processors page.
| Processor | Purpose | What it can see |
|---|---|---|
| Supabase | Database, authentication and file storage | All workspace data: deals, contacts, activity, uploaded documents |
| Vercel | Application hosting and server functions | Data in transit while a page or API call is being served |
| Anthropic | AI drafting, summarising (including meeting transcripts you paste) and research, where you ask for it, and tidying a contact spreadsheet you import | The project context for the drafts, summaries and research you ask for, a mandate letter you ask it to read for its terms, and a meeting transcript you paste to turn into notes. The rows of a contact spreadsheet you import into the Investor log, every column as it is in the sheet, which it tidies - for a sheet of up to 300 rows; a longer sheet is read by fixed rules, with no AI. Not used to train models. |
| TypeSafe AI | Reading email from your project contacts, and replies to your outreach from their colleagues at the same firm written from the firm's own company domain, to set each contact's status (the Jev model), only on projects where automatic reply status is switched on | On those projects only: the subject and text of each email one of the project's contacts sent to your connected mailbox in the last 14 days, other than those recognised as automatic replies, with the names of its attachments, the project codename, the client's name and that contact's current status. Never the attachments themselves, and no mail from anyone who is not a contact on the project, except a reply to your outreach from someone at one of the project's firms, written from that firm's own company domain, which is read as that firm's reply. E-signature notices and the mail you sent, read on those projects to record a signed NDA or a CIM, are read on our own server by fixed rules and are not sent to TypeSafe AI. Its terms bar training on this text without our consent, which we have not given, but let it use telemetry from it - logs, classifications, metrics and learnings - without restriction. |
| RocketReach | Finding people at a firm and their work address, where you ask for it | Only the firm name, and the person's name or role when you give one, for that one search. No deal data, no contact list. |
| Anymail Finder | Finding a named person's work address, where you ask for it | Only the person's name and the company's domain or name, for that one search. No deal data, no contact list, and nothing that says which customer or mandate asked. Its terms let it use search data for its own commercial purposes, keep each search for up to 12 months, and, on request, tell a searched person that our account searched for them. |
| Resend | Our own transactional email: team invites, access decisions, document alerts and reply alerts | The recipient's name and email address and the text of that one message. A document alert names the counterparty who opened a tracked document or accepted your confidentiality wording, their firm, the document type and the time - never the mandate, the client or the file name. A reply alert names the counterparty who replied, their firm and the time, and an alert that an investor has gone quiet names them, their firm and how many days - never the mandate, the client or the text of the email. Never your outreach, which does not pass through us at all. |
| Stripe | Payment processing, if and when you subscribe | Billing details only. No deal data. |
What we do not have yet
A reviewer will find these anyway, so here they are first.
- No SOC 2 Type II or ISO 27001 certification of our own.
- No formal penetration-test report to share yet.
- A hardened, tamper-evident audit log with a dedicated actor field and login, export and permission events is in progress.
- No published uptime SLA. We do not run one for a product at this stage, and we will not pretend to.
- No zero-data-retention agreement with TypeSafe AI, which reads email from project contacts, and replies to your outreach from their colleagues at the same firm, on projects where automatic reply status is switched on. Under its standard data processing terms TypeSafe keeps what it is sent for as long as necessary for the processing, and may use telemetry from it without restriction. It offers zero retention to enterprise customers; we have not arranged it.
- No data processing restriction with Anymail Finder beyond its standard terms, which let it reuse search data, keep each search for up to 12 months and tell a searched person, on request, that our account searched for them.
Something missing that your firm needs answered? Write to hello@deallaunch.ai and the answer gets added to this page.