Midjourney Prepares Web Tool for AI Image Editing Amid Deepfake Concerns
Midjourney plans to launch a web-based tool that allows users to edit any uploaded image using its generative AI. The company rolls out the feature with initial access restrictions and increased moderation to address growing concerns over deepfakes and copyright infringement.
Midjourney prepares to release an upgraded web tool that lets users edit any uploaded image using its generative AI technology. CEO David Holz states that the feature will launch early next week and includes the ability to retexture objects in images by repainting their colors and details based on text captions. The company initially restricts access to a subset of its current community as it seeks to gauge user feedback and manage potential risks.
The introduction of this editing capability highlights ongoing industry challenges regarding AI transparency and image provenance. While Midjourney previously commits to embedding basic metadata to denote AI-generated content, it notably avoids adopting the more comprehensive C2PA standard that traces an image's full creation history. This lack of robust provenance tracking raises questions about how edited images will be identified as they circulate across the internet.
Releasing such a powerful editing tool carries significant risks, including massive copyright infringement and the accelerated spread of misleading deepfakes. Concerns are mounting as deepfake creation surges by 900% compared to last year, with recent fake images of Hurricane Helene's aftermath highlighting the real-world impact of this technology. Although more than ten U.S. states enact laws against AI-aided impersonation, Midjourney relies on a combination of human and advanced AI moderators to prevent abuse of its new feature.