X reinforces regulations on AI: jailbreak bans and controversy over data

The changes that X will implement starting January 15, 2026, mark a turning point in how the platform handles AI-generated content. The update to its terms of service significantly expands what is considered “user content,” now explicitly including (prompts) and AI model outputs.

Broad License on AI Data

One of the most relevant provisions is that X grants itself a worldwide, perpetual, royalty-free license to use all this data for training AI systems and other commercial purposes. This clause fundamentally transforms the value the platform derives from users’ interactions with AI technology, solidifying its position as a data provider for model development.

Anti-jailbreak Measures and Their Implications

The new terms introduce specific rules aimed at combating techniques to bypass AI systems, particularly prompt injection and other jailbreak methods designed to evade security restrictions. X states that attempts to breach these controls will constitute contractual violations, sending a clear signal about its focus on protecting its systems.

Financial Penalties and Dispute Resolution

To reinforce compliance, X establishes a fine of $15,000 for unauthorized data extraction. Additionally, all disputes arising from these terms will be resolved in Texas courts, changing previous resolution mechanisms and concentrating jurisdiction in a specific location.

Reactions and Concerns

The update has sparked criticism from multiple fronts. AI researchers and academics express concerns about how these restrictions could limit legitimate AI safety research. Meanwhile, privacy activists question whether users truly understand how their content will be monetized through AI training.

X also reaffirms its commitment to anti-money laundering and counter-terrorism financing regulations, aspects that remain priorities within its compliance framework.