What’s happening today?

Over the past year, diffusion and GAN-based models like Stable Diffusion, DALL·E 3, and Midjourney have leapt in realism, producing imagery near indistinguishable from photographs and enabling applications from rapid prototyping in industrial design to augmented medical imaging for diagnostics.

Yet, researchers warn that deepfake detectors lag behind these generative engines, with major vulnerabilities discovered in widely used detection tools – raising concerns that synthetic images of private individuals can proliferate unchecked. Meanwhile, platforms such as Instagram are rolling out AI features (“Movie Gen”) that let any user create face-swapped videos – prompting civil-society calls for stricter transparency and labeling mandates.

Legal scholars and policymakers are scrambling: the U.S. Copyright Office’s July 2024 report urged new federal legislation (e.g. the NO FAKES Act) to give individuals enforceable rights over their likenesses, even as states and the EU update privacy laws to cover biometric data and unauthorized synthetic media. 

This duality – unprecedented creative potential versus a privacy and trust abyss – frames today’s clash over image synthesis.

At the center of this debate are Emad Mostaque, co-founder of Stability AI and architect of the open-source Stable Diffusion model, and Danielle Citron, previously the Executive Director of the Cyber Civil Rights Initiative and Professor of Law at Boston University. She now serves as Jefferson Scholars Foundation Schenck Distinguished Professor in Law at the University of Virginia School of Law.

What does AI think?