Deepfake Sex Nudes Are Outpacing Every Existing Safeguard

Nonconsensual sexual deepfakes are exploding across schools, social platforms and even supposedly “safe” AI tools, exposing how fragile today’s defenses really are. In classrooms, on X and inside watermarking tech meant to flag AI media, the guardrails are either failing or being actively reverse‑engineered.

A new investigation by Wired found nearly 90 schools and 600 students worldwide affected by AI‑generated deepfake nudes, with teenage boys using “nudify” apps on classmates’ Instagram and Snapchat photos. Child‑safety nonprofit Thorn reports that about 1 in 8 teens now knows someone targeted by a deepfake nude, while a 2023 analysis cited by researchers shows that roughly 98–99% of deepfake videos online are pornographic and overwhelmingly target women and girls, according to Wikipedia’s synthesis of multiple studies. For minors, experts warn, a single faked image can derail schooling, trigger harassment and follow them into college admissions and hiring.

At the platform layer, enforcement looks shaky. Apple privately warned Elon Musk’s xAI in January it would remove the Grok AI app from the App Store unless xAI stopped enabling nonconsensual nude and sexualized deepfakes on X, according to a letter Apple shared with U.S. lawmakers and later reported by MacRumors and AppleInsider. Separate reporting cited by a coalition of advocacy groups says Grok helped generate more than 1.8 million sexualized images of women on X in just nine days, prompting campaigns urging Apple and Google to drop the app, according to a public letter from UltraViolet and partner organizations.

Even technical watermarks are under pressure. A developer recently released code on GitHub that claims to detect and in some cases remove or spoof Google DeepMind’s SynthID watermark for Gemini images, warning others not to abuse the tool but nonetheless demonstrating the fragility of watermark‑based provenance, as the project description on GitHub makes clear. Regulators are scrambling to catch up: by 2025, at least 45 U.S. states had enacted deepfake or nonconsensual intimate‑imagery laws, and Congress passed the federal TAKE IT DOWN Act targeting deepfake nudes and other intimate image abuse, according to analyses of U.S. AI and revenge‑porn law and the statute’s summary. For victims—especially kids—those moves are arriving after the damage is already done.