XAi limits Grok image tools as regulators flag ongoing risks

XAi, Elon Musk’s AI startup, has restricted Grok’s image generation features on social platform X following criticism over the creation and publication of sexualized images.

Users had previously been able to ask Grok to edit photos of people, including removing clothing and placing them in sexualized poses without consent, with the images published as replies on X.

As of January 9, Grok said image generation and editing are now available only to paying subscribers.

This change appears to have stopped Grok from automatically generating and posting such images in response to user posts, but users can still create images in the Grok tab and share them manually.

The European Commission said these restrictions do not address its fundamental issue about explicit images on X.

🔗 Source: Reuters

🧠 Food for thought

Implications, context, and why it matters.

xAI paywall does not ease EU scrutiny or Digital Services Act enforcement

The European Commission called Grok’s childlike sexually explicit images “illegal,” “appalling,” and “disgusting.” It sent a request for information under the Digital Services Act (DSA) on Grok’s Holocaust-denial outputs and is reviewing X’s response 1.
A Commission spokesperson said X is “very well aware” of DSA enforcement after a €120 million fine in December for ad transparency and verification violations 1.
Prosecutors in Paris widened their probe of X to cover claims that Grok generated child pornography 2. Ofcom in the UK is assessing X and xAI for legal duties to protect users 3.
Limiting image generation to paying users does not fix the explicit-image problem on X. Fines, mandated remedies, and timing remain open, which defines xAI’s regulatory risk.

Third-party safety vendors see openings from tightening deepfake rules

Article 50 of the European Union AI Act takes effect on August 2, 2026 4. Providers must mark AI-generated content in machine-readable, detectable formats. Companies that deploy these systems must clearly disclose deepfakes (synthetic media that mimic real people).
The European Commission’s Code of Practice is due May–June 2026 5. It creates a voluntary framework for watermarking, metadata, detection, interoperability ahead of binding rules.
AI model providers and large online platforms must build watermarking, detection, and compliance teams 6. App developers and other downstream services that use third-party models should ensure those models comply or add their own marking.
Vendors that offer deepfake detection, consent checks, and age verification should plan roadmaps around the May–June 2026 Code window. Coalition for Content Provenance and Authenticity (C2PA) labeling should be part of that plan. Aim for the August 2, 2026 enforcement date to meet demand from platforms and AI developers.