Legal
Responsible AI policy
How Montr Business Intelligence (Montr BI) keeps AI under your control, with approvals before sends, calls and spend, paused ads, hard limits, and outside text treated as data.
Draft for review. This document is being finalised with legal counsel before Montr BI’s public launch. Questions: hi@montr.io
Our approach
Montr Business Intelligence (“Montr BI”, operated by Cloud Fold Studio) uses AI to do work for you. Our rule is simple: the AI can prepare things freely, but anything that reaches your customers or costs money waits for a person. This page describes the guardrails that exist in the product today. It is a description of how Montr works, not a guarantee of any outcome.
Approvals before anything risky
- Every agent action that sends, publishes or spends can wait for a human.
- You choose how much freedom the agent has, from asking about everything to running on its own for low-risk work.
- Some actions always ask, whatever you choose: WhatsApp sends, phone calls, campaign sends, destructive CRM actions such as deletes, and paid image, video or audio generation.
- Approvals show you what will happen before it happens. They can be delegated, they expire after a default 24 hours, and you can decide them from the desktop app, the phone app or, for approved phones, WhatsApp. Ad and bulk-call approvals stay in the app.
- Actions that move money need a second approver, so no one person can approve spend alone.
- AI tools that connect through the API or the MCP server can never approve their own actions or spend money.
Ads are created paused
When Montr creates an ad campaign, it creates it paused. Montr has no way to switch a campaign on, change it or delete it in Google Ads or Meta Ads. You turn campaigns on in the ad platform yourself. Creating an ad from the agent needs an approval card that shows the full draft, and every ad write is logged. The one exception is opt-in conversion reporting, which is off by default.
Hard limits on what the agent can do
- Each mission has hard limits on tool calls, tokens, credits and time. A run stops before the step that would go over its limit.
- You can pause or stop a mission, and a stop takes effect at the next tool call.
- Credits are reserved before an AI call and settled after. Your organisation can set spend caps, and low-balance alerts exist.
- When an agent runs a script, it runs in a short-lived sandbox with no network, files or environment access, and scripts are refused on unattended runs.
- Montr’s fetches on your behalf go through a guard that blocks internal and private network addresses.
Outside text is data, not orders
Emails, messages, web pages and documents the agent reads come from outside Montr. The agent treats that text as information to read, not as instructions to follow. This reduces, but cannot remove, the risk that a hostile message tries to steer the agent, which is another reason the approval steps above exist.
Respecting your contacts
- Opt-outs and unsubscribes are applied before sends. Do-not-contact checks run before WhatsApp, SMS and voice sends.
- WhatsApp template and opt-in rules are enforced.
- You can export or erase everything held about one contact.
Your data and AI
When you use an AI feature, the prompt and the content needed for the request go to Montr’s service and on to the AI model provider that handles it. The providers are listed on the Subprocessors page. [AI-training position — to be confirmed with legal counsel: whether Montr or its model providers may use your content to train models.] Credentials you store are encrypted at rest.
What AI can get wrong
AI can produce wrong, biased or out-of-date output, and can sound confident when it is wrong. Review anything that goes to a customer, and do not rely on AI as the only basis for decisions that seriously affect a person. Montr is in early access, and most modules are labelled Beta.
Your responsibilities
You are responsible for the settings you choose, the content you approve, and following the Acceptable use policy and the law that applies to you, including telling people when they are talking to an AI where the law or channel rules require it.
Reporting a concern
If Montr’s AI behaves unsafely or you find a way around a guardrail, email hi@montr.io. [Security and AI-safety reporting address — to be confirmed.]
Contact
Email hi@montr.io. Entity: [Legal entity name — to be confirmed].