Legal
Acceptable use
On this page
In short: The short version: do not use what we build to message people who did not ask, to deceive anyone about who they are dealing with, or to do anything illegal. We would rather lose the work than help with any of it, and we have written down what we will not do so nobody wastes a meeting finding out.
1. Who this applies to
Everyone who uses something we built or operate: clients, their staff, and anyone acting on their behalf. It is part of the terms of service.
2. Messaging — the rules that actually get enforced
These are not our preferences. Carriers enforce them, and a violation can get a phone number blocked within hours and a business barred from A2P messaging entirely.
- Consent before the first message. Every number must have opted in, with a record of the wording they saw and when. No purchased lists, no scraped numbers, no “they gave us the number for something else”.
- No pre-ticked boxes. A consent checkbox that is checked by default is not consent, and it is an automatic campaign rejection.
- STOP works immediately, in every campaign, without exception or a retention attempt.
- Identify the sender in the first message of any conversation.
- No SHAFT content — sex, hate, alcohol, firearms, tobacco — and none of the other categories carriers prohibit outright: cannabis (including CBD), payday and high-interest lending, debt relief and credit repair, gambling, multi-level marketing, and “get rich quick” offers.
- No URL shorteners on shared domains. They are treated as a spam signal and will get messages filtered.
We build the consent record and the opt-out into every system we ship. A client who asks us to route around either is asking us to end the engagement.
3. Email
The same principle, under CAN-SPAM: no false headers or misleading subject lines, a real physical address in commercial mail, a working unsubscribe honored within ten business days, and no sending to harvested addresses.
4. What we will not build
Naming these in advance so a proposal conversation does not get three weeks in before it stops.
- Anything designed to deceive someone about who they are dealing with — an AI presented as a specific named human, a cloned voice used without that person’s written permission, a fake review, a fabricated testimonial, an invented credential or partnership.
- Systems that score or rank individuals, infer protected characteristics, or show one person a different price from another based on personal data.
- Automated decisions with a legal or similarly significant effect on a person — credit, employment, housing, insurance, benefits.
- Scraping a site in breach of its terms, or any system whose purpose is to collect personal data people did not knowingly provide.
- Anything intended to harass, stalk, defraud, or discriminate against a person or a protected class.
5. Technical
- Do not attempt to access another tenant’s data, probe for vulnerabilities without telling us, or test the rate limits to see where they are. If you want to test security, tell us first and we will arrange it — see the security page.
- Do not resell access to an AI operation as a standalone product. Using Erica to run your business is the deal; reselling the model behind it is not.
- Do not upload malware, or content that infringes somebody else’s copyright or trademark.
6. Using the AI features
The AI use policy sets out what the models do. The obligation on your side is narrower: do not disable the human approval step on anything that will be published under your name or sent to your customers, and do not feed us personal data you had no right to collect.
You remain responsible for what goes out under your name. A model drafted it; you published it.
7. What happens if this is broken
In most cases we will call you, explain the problem and give you a chance to fix it — the usual cause is a misunderstanding rather than bad intent.
Where the breach is putting people at risk, breaking the law, or threatening a carrier registration or ad account that other clients also depend on, we will suspend the affected function first and talk afterwards. Repeated or deliberate breaches end the engagement, and you keep what has already been built and paid for.
8. Telling us about a problem
If something built by us is being used in a way this policy prohibits, email admin@inericasystems.com. We read it, and we would rather hear about it early.