Feiyou Guo, who is building monid.ai as a tool layer for agents, said X suspended both his account and the project's account without an alert or apparent avenue for recourse. In a post showing the suspensions, Guo expressed astonishment at the lack of notice and a clear path to challenge the action.
The episode highlights a basic risk for founders and developers who depend on X: The platform can restrict access while providing too little information to determine what triggered the decision or how to reverse it. Guo's screenshot establishes that both accounts were suspended, but it does not show why X acted.
A separate test by Ryan Merket provides supporting evidence of the same opacity on content distribution. Merket prompted Grok to audit RuntimeWire's X account and assess whether the platform had algorithmically limited its reach. In a screenshot shared by RuntimeWire, Grok assigned an approximately 80% likelihood that an algorithmic filter was responsible, citing low views, engagement velocity, quality scores and outbound-link penalties.
The estimate cannot establish that X filtered the account. The screenshot includes no account analytics export, historical impression baseline, comparable control accounts or ranking logs. Without those records, Grok cannot distinguish deliberate filtering from changes in audience behavior, distribution or other platform systems. Merket's test instead reinforces the problem raised by the suspensions of Guo's and monid.ai's accounts: Even the platform's own assistant cannot provide the internal evidence needed to explain a loss of reach or access.
The uncertainty matters to founders building AI agents on social platforms. X permits automation while retaining broad discretion to restrict API access, filter content and suspend accounts. Developers can comply with published rules and still lack the internal data needed to explain a sudden enforcement action.
X permits automation, with broad enforcement discretion
X's automation policy, updated in April 2026, permits developers to build services that automatically publish useful information, respond to users and operate through direct messages.
The same policy prohibits spam, substantially similar posts across multiple accounts, non-API browser scripting, rate-limit circumvention, misleading links and automated actions taken without sufficient user consent. X says violations can lead to search filtering, API restrictions or account suspension.
Qualifying automated accounts must also carry labels and remain connected to a human-run account. X's developer policy says commercial products that move beyond prototyping or a limited number of users must apply for an enterprise plan. The company can monitor API use, require developers to keep approved use-case descriptions current and revoke access when it determines that a product has departed from those descriptions.
Rapid follower or engagement growth is not identified as a violation in the published policies. A fast-growing autonomous account could still produce behavior that X categorizes as spam, manipulation or unauthorized automation, leaving its developers dependent on enforcement systems they cannot inspect.
The ElizaOS dispute documents the platform dependency
Federal court records provide a documented example of an AI-agent developer colliding with X, although the case ended without a ruling on the underlying allegations.
X suspended accounts belonging to ElizaOS, an open-source framework for autonomous agents, and its founder, Shaw Walters (@shawmakesmagic), around June 10, 2025. Eliza Labs later bought an enterprise license from X after suing the platform and xAI, according to an X lawyer quoted by Reuters.
Walters founded Eliza Labs in 2024 after working on games, digital characters and crypto projects. He and 13 co-authors described the framework in a January 2025 paper as open-source, TypeScript-based software that developers controlled themselves.
The ElizaOS repository now describes the project as a local-first operating system for agents, with browser automation, voice, messaging, wallets and optional cloud services. GitHub showed roughly 5,600 forks when reviewed. The framework supports X alongside Discord, Telegram, WhatsApp, Slack and Farcaster.
Eliza Labs and Walters filed their original complaint on August 27, 2025. An October 27 transfer order summarized their allegations without deciding whether they were true.
The plaintiffs claimed X requested information about how ElizaOS agents operated and later used the resulting technical explanations to improve competing xAI products. They also alleged X demanded an enterprise API license costing $50,000 a month, or $600,000 a year, before suspending the @ElizaOS and @shawmakesmagic accounts and blocking links to the project's website and GitHub repository.
X denied wrongdoing and maintained in court that the accounts had violated its terms. The California court transferred the case to the Northern District of Texas because Walters and Eliza Labs had accepted X agreements requiring disputes to be heard there.
A federal judge dismissed the case with prejudice on December 29, 2025, after the parties agreed to end it. Reuters reported that X's lawyer said Eliza Labs had purchased an enterprise license and that neither X nor xAI paid the company to resolve the dispute.
The dismissal did not establish that X copied ElizaOS, selectively enforced its rules or used API pricing to constrain a competitor. It also did not produce a public explanation of the suspensions.
Opaque enforcement compounds platform dependency
The suspensions of Guo's and monid.ai's accounts put the lack of notice and apparent recourse at the center of the platform risk. Merket's test shows a related limitation: Grok is embedded in the platform whose distribution decisions it was asked to assess, but its estimate did not include the internal evidence required to verify its conclusion.
xAI has described real-time access to X as a fundamental advantage for Grok and distributes the assistant through X. The platform is therefore a distribution channel, data source and enforcement authority for outside agent developers while Grok competes across automated search, publishing and assistant use cases.
Walters continued working on ElizaOS after the lawsuit. In March 2026, he said the project remained under development and pointed to its plugins, games and planned agent products. The repository now emphasizes software that can run locally rather than depending entirely on social platforms.
That architecture reduces one form of platform risk but cannot eliminate the distribution problem. Open-source code protects a founder's ability to keep building. It does not guarantee access to the proprietary networks where an agent finds users, data and attention. The suspensions of Guo's and monid.ai's accounts are the main example of that exposure; Merket's Grok audit and the ElizaOS dispute provide supporting evidence that founders may lack both the internal records and meaningful recourse needed to determine why access or distribution changed.