Tests Find Grok Still Generates Sexualized Images Despite xAI’s Claimed Safeguards

Overview

• An O Globo test of the Grok iOS app added bikinis and lingerie to women and, in one instance, placed swimwear on a child image despite xAI’s recent claim that it blocks edits depicting real people in revealing clothing.
• Investigators in Brazil—ANPD, MPF and Senacon—issued formal recommendations requiring X to prevent new sexualized images, remove existing ones, suspend offending accounts and deliver monthly compliance reports.
• Independent analyses reported large‑scale production of suggestive content on Grok, with AI Forensics finding that over half of 20,000 reviewed images showed scant clothing, 81% depicted women and about 2% appeared to involve minors.
• xAI and Elon Musk said the system refuses illegal requests and attributed issues to adversarial prompts, while X first restricted image tools to paying users, then widened limits, though premium subscribers can still edit compliant AI‑generated images.
• Several jurisdictions have investigated or restricted Grok, including Indonesia and Malaysia, and Brazilian law criminalizes creating or sharing manipulated sexual images, with harsher penalties for dissemination and for cases involving minors.