Trust center

How we keep you in control

What it will never do, how your yes works, where your data lives and what we share. Most lines link the rule that enforces them.

The check

Before anything goes out

  • Every post, email and reply passes one check before it leaves. It looks up what you allowed for this brand, then lets that one action through once.
  • Spending money, opening accounts and deleting data are refused at every setting. Nobody can loosen that, us included.
  • Your yes covers the exact words. If anything changed since you looked, it refuses and asks again.
  1. 01An actionPublish a post, send a batch
  2. 02The checkWhat you allowed for this brand
  3. 03One passGood once, for these exact words
  4. 04It goesOr waits for you, or is refused
Floors that no setting can lower
ActionAt every setting
Spend moneyrefused
Create an accountrefused
Delete datarefused

Your data

Data handling

  • Each company's work runs on its own, with its own files and its own locked keys for your accounts. One company can never read another's.
  • The record of what you approved and what went out cannot be edited or deleted by anyone, the owner included.
Record of what happened, sample rows
TimeWhoWhatWords
09:30:12ownerapproved postversion 1
09:31:40owneredited post, asked againversion 2
09:32:05Kirtispublishedversion 2
Sample rows. Nobody can change or delete a row, the owner included.

Shared numbers

The shared record, and how to leave it

  • We pool numeric outcomes across workspaces: investor reply and meeting rates per firm and stage, cold email reply rates per segment, post performance per hook pattern. No content and no personal data.
  • An aggregate is published only when at least 10 workspaces contribute to it, at firm level, never per partner name.
  • It is on by default. Turn it off in Settings, under Outcome commons; your workspace stops adding to it from that moment.
  • Replies from seed funds, for money apps7 of 10 companies, hidden
  • Meetings with angels, for software firms14 of 10 companies, shown
Sample rows. The mark is where a row starts to show: 10 companies.

Labels

AI disclosure

  • AI hosts and AI voices carry an AI-made label, and a host needs two approvals before it appears.
  • Article 50 of the EU AI Act applies from 2 Aug 2026. Posts with an AI host or AI voice carry a visible label and the network's AI flag; machine-readable marks on generated images and video are not built yet.
  • We do not make synthetic customers, reviews or testimonials.

Northwind Labs · 0:42 video

Host: AI character, approved twice. Voice: AI voice.

Sample post. The label rides on the media, not only in the caption.

Never

What we refuse to do

  • Post, reply or email without your approval of the exact content, for fresh public copy.
  • Automate engagement on networks that forbid it; those become manual replies with the text ready.
  • Email anyone who opted out, from any brand or from fundraising.
  • Write a number into your copy that you did not confirm.
Refused, every timeall settings
  • Fresh post without approvalrefused
  • Automated engagement where bannedrefused
  • Email to a suppressed addressrefused
  • A number you did not confirmrefused

each line is a fixed rule, not a setting

Subprocessors

Who processes data for us

Subprocessors
CompanyPurposeRegion
Google CloudHosting, database, storage, key management; Gemini and Veo models on Vertex AIEU and US
AnthropicClaude models for planning and drafting, through Vertex AI or from Anthropic directlyUS
StripeBilling and, if you connect it, revenue readsUS and EU
PostmarkSign-in and notification emailUS

Channel providers (Meta, LinkedIn, TikTok, Google, X, Reddit, Telegram, Discord) receive only what you approve to publish, under your own accounts.