The scale of AI image generation in 2026 is difficult to overstate. Midjourney alone reports generating over 15 million images per day. When combined with Stable Diffusion (which runs on private hardware with no usage reporting), DALL-E, Adobe Firefly, and thousands of smaller models, the global output of AI-generated images exceeds that of all camera photographs produced daily. Within this volume, non-consensual intimate imagery (NCII) represents a growing and legally significant category.

Volume of AI-generated images

Industry estimates for 2026 put global AI image generation at 15 billion or more images per month across all platforms and local installations. Midjourney processes over 15 million image generation requests per day. DALL-E through ChatGPT generates hundreds of millions more. Open-source tools (Stable Diffusion and derivatives) run on an estimated 15 to 20 million personal computers and generate an unknown but large additional volume. For context, approximately 1.8 trillion photographs were taken in 2023, roughly 5 billion per day. AI image output is now a significant fraction of that total.

Deepfake NCII statistics

Non-consensual intimate imagery (NCII), including deepfake pornography, represents approximately 90 to 96% of all deepfake content online by volume, according to multiple research reports from 2024 and 2025. Women are the target of over 99% of deepfake NCII. The average deepfake NCII victim discovers the content 14 months after it was first created. Removal without legal assistance takes an average of 18 months. With TAKE IT DOWN Act notices, the mandatory platform removal window is 48 hours.

Detection accuracy statistics

Expert human reviewers correctly identify AI-generated images approximately 30 to 40% of the time in blind tests using images from 2025 and 2026 generation models. The best-performing ML-based detection systems achieve 85 to 95% accuracy on benchmark datasets but significantly lower accuracy on adversarially post-processed images. EXIF-based metadata detection achieves over 90% accuracy on unmodified AI outputs because AI generators do not embed camera metadata. Dimension-based detection achieves 80%+ accuracy on images at standard generator output sizes.

Legal and platform response statistics

The TAKE IT DOWN Act (2026) imposes a 48-hour removal obligation on covered platforms upon receipt of a compliant notice. Major platforms (Meta, LinkedIn, TikTok, Snapchat) tend to act quickly; Reddit and X tend to use more of the statutory window; imageboards and Telegram require escalation more often. The legislation covers all 31 jurisdictions in which ScanErase operates. 26 US states have enacted additional state-level deepfake NCII laws as of 2026, with criminal penalties in 18 of those states.

Frequently asked questions