Last updated: October 28, 2026

Are you responsible for what your AI agent posts on X?

Yes. Since October 9, 2026, X's Terms of Service say you are responsible for your use of X "including any features that perform autonomous actions", and that the law applies to "autonomous actions on your behalf". Whatever an AI agent posts, replies, likes or sends from your account is treated as you doing it.

Letting an agent post is still allowed. The safe setup is the one where the agent writes and a human, or a schedule and a cap you chose, decides what goes out. This is a plain reading of the published terms, not legal advice.

What the October 9, 2026 terms say

Quoted from the X Terms of Service, read on October 28, 2026.

Clause
The text
What it means for agents
Section 3, your responsibility
"You are responsible for your use of the Services, including any features that perform autonomous actions"
An agent acting through X counts as you acting.
Section 3, content
"including any inputs, prompts, outputs, and/or information obtained or created through the Services"
What the AI writes is your content, and so is the prompt.
Section 3, the law
compliance with applicable laws covers "autonomous actions on your behalf"
Defamation, financial claims or election rules apply to agent posts as if you typed them.
Section 4, misuse
"scripting the creation of Content in such a manner as to interfere with or create an undue burden on the Services"
High-volume scripted posting can break the terms on its own.
Section 4, scraping
"crawling or scraping the Services in any form, for any purpose without our prior written consent is expressly prohibited"
Agents that read X by browser automation are out of bounds.

What changed, and why now

X announced the revision on September 9, 2026, and it took effect on October 9. According to a clause-by-clause comparison with the previous terms, the old text already made you responsible for prompts and outputs. The new part is autonomy: features that act without you pressing a button, and actions taken on your behalf.

It lands as agents that act for you go mainstream: Grok bots on X itself, Claude Duet, ChatGPT dots and Meta Muse launched between August and October 2026. The same revision adds a jury-trial waiver where the law permits, arbitration if the Texas court clause cannot be enforced, and names SpaceXAI, Cursor and SpaceX among X's affiliates. Using X on or after October 9 counts as accepting the terms.

"The AI did it" is not a defence

If your agent posts a false claim about a company, a fake price target or a duplicate post 50 times, the account that gets limited, and the person answerable for it, is you. The terms put that in writing.

What your agent does, and what can go wrong

X does not publish a list of banned agent behaviours. This is how its automation rules and the new terms apply to the things agents are usually asked to do.

Agent action
Risk to your account
Do this instead
Posts to your own timeline
Low, if each post is original
Vary content; never post the same text twice
Replies to strangers or mentions them
High: the most reported automation offence
Only reply when someone asked; approve each reply
Likes, follows, reposts on its own
High
Do not automate these at all
Sends DMs
High without consent
Only to people who opted in, approved by a human
Posts links in bulk
Medium: spam signals and link fees
Cap link posts per day
Reads X by scraping
Banned outright; liquidated damages over 1M posts a day
Read through the API or licensed endpoints

The scraping clause comes with liquidated damages of $15,000 per 1,000,000 posts for requesting, viewing or accessing more than 1,000,000 posts in any 24-hour period (Section 5).

A checklist before you connect an agent

  1. Decide what it may do: post to your timeline only. No automated likes, follows, reposts or unsolicited replies.
  2. Start in draft mode. Let the agent fill a queue you approve, and switch to auto-publishing only after a few weeks of drafts you did not have to fix.
  3. Cap it: a daily post limit and a daily link limit, set somewhere the agent cannot change.
  4. Make it cite its sources for any fact, number or quote, and check them.
  5. Turn on the Automated label in your X account settings if the account is mostly automated.
  6. Keep a log of what it posted and why. If X asks, or a post goes wrong, you can show what happened.
  7. Never let it read X by scraping or a logged-in browser session.

How OpenTweet keeps you in control

OpenTweet is the posting layer between your agent and X. Connect Claude, ChatGPT, Cursor or any agent through the MCP server or the REST API, and it posts through OpenTweet's approved X app, so there is no developer account or scraping involved.

  • Drafts first. A post sent without a time or publish_now is saved as a draft for you to review in your queue.
  • Caps the agent cannot raise. Daily post and link-post limits are set by your plan, not by the agent.
  • Repeat guard. The API stops a post that repeats one of yours unless it is resent with allow_duplicate.
  • Timeline only. The MCP tools draft, schedule and publish posts, threads and articles on your own account. There are no tools for liking, following or replying to other people.

7-day free trial with a few limits, or pay now for full limits.

Frequently asked questions

Who is responsible if my AI agent posts something on X?

You are. X's Terms of Service effective October 9, 2026 say you are responsible for your use of the Services, "including any features that perform autonomous actions", and that complying with the law covers "autonomous actions on your behalf". If an agent you connected posts, replies, likes or DMs, X treats it as you doing it.

What changed in the X Terms on October 9, 2026?

X announced the revision on September 9, 2026. The earlier terms already made you responsible for "inputs, prompts, outputs" from AI features; the new text adds features that act on their own and actions taken on your behalf. It also adds a jury-trial waiver where the law allows it and arbitration as a fallback if the Texas court clause is unenforceable. The AI-training license on your content was already there before.

Does that only apply to Grok?

The text does not name Grok or any product. It covers "any features that perform autonomous actions" through the Services. A third-party agent posting through the X API is also governed by the X Developer Agreement and automation rules, and the account it posts from is still yours.

Is it allowed to let an AI agent post to my X account?

Yes, through the official API or an approved app. What gets accounts limited is the behaviour: posting the same or near-duplicate content, automated replies or mentions to people who did not ask, automated likes and follows, and spam links. Automated accounts should also turn on the Automated label in their X settings.

Can my agent read or scrape X for research?

Not by scraping. The terms say crawling or scraping "in any form, for any purpose without our prior written consent is expressly prohibited", and set liquidated damages of $15,000 per 1,000,000 posts for requesting, viewing or accessing more than 1,000,000 posts in any 24-hour period. Use the API or a tool that reads through licensed endpoints.

What is the safest way to run an AI agent on X?

Have the agent write drafts, not posts. With OpenTweet, an agent connected through the MCP server or the API saves drafts to your queue and you approve them, or you let it publish only on a schedule you set, with daily post caps per plan. It posts through OpenTweet's approved X app, so you need no developer account. Plans start at $19.99/mo.

Keep exploring

X rules for automation, AI content and agents.

X automation rules

What automated accounts may and may not do on X in 2026.

X API rules for AI agents

The developer-side rules for agents that post through the API.

Does X allow AI-generated content?

Yes, with conditions. What X penalizes and what it labels.

Is it safe to connect an MCP server to X?

What an MCP server can do on your account, and how to limit it.

Which personal AI agents can post to X?

Claude Duet, ChatGPT dots, Meta Muse and Grok compared.

Did X suspend my bot in the purge?

How to tell, and how to keep an automated account safe.

Let your agent write. You decide what posts.

Drafts in a queue you approve, caps it cannot change, and the official X API underneath.

7-day free trial with a few limits, or pay now for full limits.