AI disclosure
Effective date: November 15, 2026
PostMesa uses AI for two things: suggesting wording for replies, DMs and rewrites when you are setting up an automation, and answering questions about PostMesa in the Ask Mesa help assistant. This page says exactly what that means.
What the AI does
- Drafts options for a public reply, a first DM, a follow-up DM, or a rewrite (shorter, warmer, in another language).
- Works only from text you give it: your post caption, your short brief, examples and your chosen tone.
- Respects protected phrases you set, such as slogans, disclaimers or sacred names, which it must not change.
What the AI does not do
- It never sends a message. A person on your workspace reviews, edits and approves every message before it is used.
- It does not chat with your followers on its own.
- It does not read commenters' messages in this version.
- Your content is not used by us to train AI models.
Who processes it
Reply suggestions: requests go from our servers through Cloudflare AI Gateway to Anthropic, PBC, which generates the draft. For suggestions the gateway records request metadata such as time and size, not the text. We store the version you accept, and mark it "AI-assisted" in the automation history, not the prompt.
Ask Mesa: your question goes through the same gateway to Anthropic, PBC, which writes the answer. See the next section for what is kept.
Anthropic's terms for API customers say it deletes API inputs and outputs from its systems within 30 days and does not use API data to train its models. Exceptions apply: content its safety systems flag as possibly harmful can be kept and reviewed by a person to enforce its usage policy, and data the law requires it to keep. We do not control Anthropic's systems; its own policies govern what it does.
Ask Mesa, the help assistant
- It is an AI. It answers only questions about PostMesa, from our own help pages, and it can make mistakes. It cannot see or change anything in your account.
- It does not answer general questions, and it does not discuss the technology or provider behind it.
- Your questions and its answers are kept in the AI Gateway log for up to 90 days, with emails, keys and card numbers removed and without your name, account, workspace or IP address. The only labels kept with them are app or website, the shape of the page, an error code if you asked from an error message, and your language.
- Only PostMesa staff can read the log. We read it to improve Mesa and our help pages and to review safety and abuse. We never use it for advertising, and we do not use it to train AI models.
- Our own logs and database never contain the text of your questions.
- Do not paste passwords, keys or personal data. Automatic removal does not catch everything.
- You can opt out: signed-in people turn off "Help improve Mesa by sharing my questions" in Settings (on by default); on the website, tick "Do not store this chat". We then ask the gateway not to keep that chat.
Cloudflare and Anthropic, PBC process these requests for us. See the sub-processors page.
Limits and your responsibility
AI drafts can be wrong or off-tone. You are responsible for what you approve. Each plan has a monthly allowance of suggestions because they cost money to produce.
Contact
Email hello@postmesa.com.