X has announced a forthcoming revision to its Terms of Service (ToS), which will take effect on January 15, 2026. This updated framework significantly redefines the concept of user-generated “Content” while incorporating contractual language pertinent to the operation and safeguarding of its artificial intelligence (AI) systems.
As it currently stands, the Terms of Service dated November 15, 2024, will remain operational until the new version supersedes it in 2026.
A pivotal amendment within this revision is X’s recognition of interactions facilitated by AI as “Content” for which users are held accountable, thereby broadening the scope of responsibility beyond traditional posts and materials.
Reconceptualization of Ownership and Responsibility in the Context of AI
The revised terms delineate that users bear responsibility for Content that encompasses “inputs, prompts, outputs,” as well as any information “obtained or created through the Services.” X explicitly advises users to contribute, create, or generate only such content they are willing to disseminate publicly.
In contrast, the current 2024 terms merely reference responsibility concerning “any Content you provide,” omitting explicit mention of prompts and outputs. This omission effectively distances Grok-style usage from the principal contractual lexicon.
This expanded definition operates in conjunction with an existing licensing agreement that confers extensive rights to X regarding content reuse. Specifically:
- Users grant X a global, royalty-free, sublicensable license to utilize, copy, reproduce, process, adapt, modify, publish, transmit, display, and distribute Content “for any purpose,” which includes its application in analyzing and training machine learning and AI models.
- No financial remuneration will be allocated for such uses, with access to the service deemed “sufficient compensation.” This stipulation renders the specific language concerning prompts and outputs particularly significant for users who may perceive AI interactions as distinct from public postings.
Moreover, the 2026 draft introduces a specific clause prohibiting conduct aimed at circumventing AI safeguards. The term “misuse” is defined to encompass attempts to bypass platform controls through methods such as ‘jailbreaking’, ‘prompt engineering’, or ‘injection’. Notably, this language is absent from the analogous misuse provisions articulated in the 2024 terms. This addition provides X with a contractual basis for enforcement actions against attempts to undermine AI feature safeguards rather than relying solely on product regulations or policy guidelines.
Region-Specific Considerations: European Legislative Context
The updated ToS incorporates modifications pertinent to European legislation that alter the document’s approach toward content enforcement and user appeals. Specifically:
- The summary and content regulations acknowledge that EU and UK legal frameworks may necessitate enforcement not solely against illegal content but also against material deemed “harmful” or “unsafe.”
- Examples of such harmful content include instances of bullying or humiliation, materials related to eating disorders, and discussions surrounding self-harm or suicide methodologies.
- The revised terms introduce UK-specific language detailing user rights to challenge enforcement actions in accordance with the UK Online Safety Act 2023.
Expanded Enforcement Provisions and User Liability Implications
The updated ToS maintains X’s stringent restrictions on automated access and data collection practices. This includes a liquidated damages schedule tied to extensive viewing activities:
- Crawling or scraping is expressly prohibited “in any form, for any purpose” without prior written approval from X. Access is generally confined to “published interfaces.”
- Liquidated damages are set at $15,000 for every 1 million posts requested, viewed, or accessed within any 24-hour period if violations occur at that scale.
- The 2026 draft refines related language to apply in instances where a user induces or knowingly facilitates such violations.
The provisions governing dispute resolution remain anchored in Texas while undergoing minor adjustments that may extend certain state-law timelines. Disputes must be adjudicated in federal or state courts located in Tarrant County, Texas. The 2026 text further stipulates that these forum and choice-of-law provisions are applicable to both “pending and future disputes,” irrespective of when the underlying conduct transpired. In contrast to the previous terms—which referenced specifically the U.S. District Court for the Northern District of Texas—the new version retains Tarrant County as the sole venue option but delineates time limits as one year for federal claims and two years for state claims; this replaces a singular one-year limitation previously established.
X continues to impose limitations on how users may pursue claims and what remedies they may recover upon prevailing in legal disputes. Notably:
- The agreement enshrines a class-action waiver that prohibits users from initiating claims as part of a class action or representative proceeding in various contexts.
- X’s liability is capped at $100 per covered dispute.
These provisions have garnered critique within broader discussions regarding whether these terms effectively diminish practical avenues for redress even when users allege significant harm.
Critique: Potential Chilling Effects on Research and Expression
The impending revisions have prompted substantial public backlash centered primarily on provisions that predate the upcoming draft yet persist within it. Critics have expressed concern regarding how these changes may curtail independent research activities:
- The Knight First Amendment Institute has articulated that X’s new terms “will stifle independent research,” labeling this approach as “a disturbing move that the company should reverse.”
- The Center for Countering Digital Hate announced its departure from X ahead of the imminent terms alteration, criticizing the Texas venue requirement as a strategic maneuver designed to favor certain judicial outcomes.
- The Reuters Institute for the Study of Journalism has highlighted how litigation can exert “a chilling effect” on critics of the platform.
- Concerns regarding AI training and licensing issues have been framed within consumer discourse surrounding user attrition from the platform.
| Clause | Current ToS (Nov. 15, 2024) | Future ToS (effective Jan. 15, 2026) |
|---|---|---|
| Definition of “Content” | User responsibility centered on content provided by users | Explicit inclusion of “inputs, prompts, outputs” along with information obtained or created through services |
| AI Circumvention Provisions | No explicit mention of “jailbreaking” or prompt-injection clauses | Prohibits circumvention attempts including ‘jailbreaking’, ‘prompt engineering’, or injection techniques |
| European Enforcement Framework | Lacks mention of UK Online Safety Act challenge processes in summary | Adds references to “harmful/unsafe” examples along with UK Online Safety Act 2023 redress procedures |
| U.S. Venue and Claim Timeframes | Northern District of Texas (federal) or Tarrant County (state); one-year deadline | Tarrant County federal/state courts; one year for federal claims; two years for state claims; forum provisions applicable to pending and future disputes |
| Scraping Penalties | $15,000 per 1 million posts requested/viewed/accessed within a 24-hour violation period | Same penalties apply with adjustments for facilitation based on user-induced violations |
With an effective date set for January 15, 2026, X’s contractual language categorically reclassifies prompts and generated outputs as user-generated Content within its licensing and enforcement framework. Furthermore, it incorporates prohibitions against practices such as jailbreaking and prompt injection into its list of impermissible conduct.
