X has added a new option within the image upload settings that enables users to block Grok from generating alternate versions of their uploaded media. The toggle, spotted in the X iOS app, appears to be a straightforward privacy protection. But the fine print tells a different story.
Testing by The Verge revealed the toggle prevents other users from tagging @Grok in replies to your images, but doesn't actually stop the chatbot from editing your uploaded images through other methods. Users enabling the feature may believe their photos are fully protected when in reality they've only blocked one specific interaction type.
The gap between marketing promise and actual function reflects a recurring pattern with Grok.X restricted Grok's ability to generate sexualized nude images in January and limited image generation features to paying users, yet users could still generate deepfake nudes via Grok with certain commands.When The Verge tested the updated system after those restrictions, Grok could still be prompted to produce sexualized edits by phrasing requests slightly differently; asking for a bikini might trigger a refusal, but asking for "revealing summerwear" or adjacent styling sometimes did not.
This matters becauseusers could previously reply to a post with edited versions of someone else's photo using text prompts, sometimes leading to unwanted alterations, artistic remixes without permission, or more concerning manipulations.The global backlash involved sexualized images of women and children, with bans and warnings from governments worldwide.
Regulators have noticed the pattern of incomplete fixes.British officials said recent steps were welcome but vowed to continue their probe, which could result in X getting banned in the UK, whilst California officials announced an investigation and the Philippines said it was moving to ban Grok.If X is found not to comply with the Online Safety Act, Ofcom can issue a fine of up to 10 per cent of its worldwide revenue or £18 million.
What makes this particularly problematic is the transparency question.X clearly understands users want protection from unwanted AI manipulation; the company built the toggle interface, wrote the feature description, and pushed it to the iOS app, but didn't follow through with comprehensive functionality to match the promise. That gap suggests either X did not invest adequately in the technical work, or deliberately chose a partial solution.
The photo toggle is positioned as a response to legitimate privacy concerns. Butthe control option, buried within upload settings, may end up saving X millions of dollars in penalties while providing users with a valuable—though incomplete—option. For users genuinely concerned about their images being processed by Grok, that incomplete protection is not good enough.