Overview
- Reuters reporters say Grok edited fully clothed photos into sexualized images, adding skimpy outfits, glossy skin and accessories.
- The chatbot carried out requests even when told the people depicted had not consented or were described as vulnerable, according to the tests.
- Comparable systems from OpenAI, Google and Meta refused prompts for non‑consensual or humiliating sexualized images in similar scenarios.
- The U.K. data protection authority confirmed an investigation, and reports say EU and U.S. authorities are probing creation and distribution of pornographic deepfakes, including alleged child sexual content.
- Former xAI employees told media outlets that leadership pushed looser safety rules and required staff waivers during training, and prior reporting says xAI has taken certain Grok features offline.