Guide
Yes, with limits. Your AI can read how your leads are doing today, but consent, quiet hours and STOP have to be checked before every text, so sending belongs inside a system that does the checking.
By the Virtual Closer team. Published 2026-11-03. Last updated 2026-11-03.
What changed: Published today. Checked the federal text on calling hours, consent and opt-outs, and the connector safety guidance from Anthropic, OpenAI and the MCP specification.
Before any message
Texting a lead is not the same as emailing one. For autodialed or artificial-voice telemarketing, the FCC rule requires the "prior express written consent of the called party" (47 CFR 64.1200(a)(2)). Whether a particular message needs that consent depends on what it says and how it is sent, so ask your own counsel about your leads. This guide is general information, not legal advice.
What this means for an AI: the AI should never be the thing that decides whether a lead agreed. Consent is a fact recorded on the lead, and it has to be checked before every send. In Virtual Closer, Mira texts leads who agreed to texts. Your own Claude or ChatGPT does not hold that check, which is one reason sending is not something you hand to a general chat.
Timing
The same rule bars telemarketing calls "before the hour of 8 a.m. or after 9 p.m." at the called party's location (47 CFR 64.1200(c)(1)). The test is the lead's local time, not yours. A chat assistant that does not know the lead's time zone cannot apply it. Some states set shorter windows than the federal one, so check the states you work in.
Mira's follow-up texts wait for 8am to 9pm in the lead's time zone, and shorter in stricter states. A text answering something the lead just did, such as a booking confirmation, can go out right away.
Opt-outs
A person can revoke consent by "any reasonable method", and the sender has a limit of ten business days to honor it (47 CFR 64.1200(a)(10)). Businesses must also keep an internal do-not-call list of people who ask not to be contacted (47 CFR 64.1200(d)). Do not wait for the limit: stop at once.
Email has its own rules. The FTC's guide says a commercial email needs accurate sender information, a clear statement that it is an ad, a valid postal address, and opt-outs honored within 10 business days (CAN-SPAM compliance guide).
In Virtual Closer, one STOP or do-not-contact entry blocks text, call and email to that lead. After a STOP, the one text sent is the confirmation that the lead is opted out. If you work leads in your own chat tool with no shared do-not-contact list, a STOP in one place will not reach the others.
Privacy
Designed so lead names, contact details and health answers are not sent to the AI model: your assistant sees a lead as a reference number and a link into the app.
In practice the connection answers about one account and one person. Counts under 30 leads show as fewer than 30, and other counts are rounded, so your assistant says about. To see who a lead is, you open the link in the app yourself.
Live today where Connect AI agent is switched on for your account: ask who replied, what your day looks like, your meetings, and your pipeline counts. Your assistant calls who_replied, today, appointments and pipeline_counts.
Coming soon: asking from the chat to add a note, change a stage, create a task, book or cancel a meeting, or hand Mira a text or email to send. These need account switches that are not on yet, so on Read only, and today, they refuse.
Do not paste a lead's name, number or health answers into any chat to work around this. Anything you paste goes to that chat tool, under its terms and not ours.
Connect safely
Claude's help center explains that OAuth sign-in lets Claude act for you without seeing your password, that you can revoke access, and that you can switch off individual tools (Claude Help Center). It also warns that Anthropic does not verify these services, that a server's developer can change tool behavior without notice, and that a malicious server may hide instructions meant to make Claude act in ways you did not intend. Its advice is to review tool inputs and outputs, and to reserve "Allow always" for servers you trust.
The Model Context Protocol specification says "there SHOULD always be a human in the loop" with the ability to deny tool calls (MCP Tools). OpenAI's guide tells builders to prefer "official servers hosted by the service providers themselves" and to require approval for sensitive actions (OpenAI developer docs). The protocol's security notes recommend starting with narrow, low-risk read access and widening it when needed (MCP security best practices).
Applied to leads: connect a server run by the company that holds your lead data, begin with a read-only connection, keep approval prompts on, and add write access later and one tool at a time. Virtual Closer offers a read-only address for this, and the settings page lets you choose Read only.
What you say
When the FTC announced its sweep of deceptive AI claims, FTC Chair Lina M. Khan said "Using AI tools to trick, mislead, or defraud people is illegal" (FTC press release). The FTC said AI use does not exempt a business from existing consumer protection law. For agents, the practical rule is that anything your AI drafts for a lead, such as a quote, a benefit or a promise about coverage, is still your statement. Read it before it goes out.
Before you start
Be straight about gaps
We tried to read the FCC's 2024 ruling on AI-generated voices and CTIA's messaging guidelines in their original text. The pages we could reach did not give us readable primary text, so we do not cite or link them here. If you call or text with an AI voice, or you are registering a texting campaign, read both yourself and check with counsel. We will add them once we have read the originals.
Go deeper
Sources
Further reading
We read each page above on 2026-11-03. Links are plain links with no payment from the publishers. This guide is general information, not legal advice.
Questions people ask
Not directly yet. Mira, Virtual Closer's assistant, texts leads who agreed to texts. Handing Mira a text from your own AI chat is Coming soon; today your chat reads who replied.
Designed so lead names, contact details and health answers are not sent to the AI model: your assistant sees a lead as a reference number and a link into the app.
The federal rule bars telemarketing calls before 8 a.m. or after 9 p.m. at the lead's location, and some states are stricter. Mira's follow-up texts wait for 8am to 9pm in the lead's time zone.
One STOP or do-not-contact entry blocks text, call and email to that lead. After a STOP, the one text sent is the confirmation that they are opted out.
No. It is general information. Ask your own counsel about your leads and your states.