Deepfakes and impersonation

YouTube is expanding its likeness detection feature to Hollywood celebrities. The tool searches the platform for AI deepfake content and flags it for enrolled public figures, who can track such videos or request removal, with takedowns judged against YouTube's privacy policy. Not every request will be approved.

According to NBC News, Apple privately threatened to remove Elon Musk's AI app Grok from the App Store in January over its failure to curb nonconsensual sexual deepfakes spreading on X. In a letter to US senators, Apple said it contacted the teams behind X and Grok and demanded that they create a plan to improve content moderation.

WIRED and Indicator reviewed publicly reported cases and found sexualized deepfakes created with generative AI targeting schoolchildren in at least 28 countries since 2023. The review counts about 90 schools and more than 600 pupils affected, with most alleged perpetrators teenage boys using "nudify" apps on images taken from social media. The authors note the true scale is likely much higher, since many incidents are never publicly reported.

Deepfake audio and video have become cheap and convincing enough to fool employees and executives. A 2025 Gartner survey found that 43% of cybersecurity leaders had experienced at least one audio deepfake and 37% had experienced video deepfakes in the past year. The article also describes a case at Arup where a roughly $25 million transfer followed a video call with AI-simulated executives.

The authors introduce DFREC, a scheme that recovers both the source and target faces from a deepfake image to support identity tracing in forensic investigations. It uses an Identity Segmentation Module, a Source Identity Reconstruction Module and a Target Identity Reconstruction Module, the last of which relies on a Masked Autoencoder. Evaluated on FaceForensics++, CelebaMegaFS, FFHQ-E4S and Celeb-DFv2, DFREC outperforms state-of-the-art deepfake recovery algorithms and is described as the only scheme that recovers both pristine faces with high fidelity.

YouTube is rolling out an AI feature in Shorts that lets creators make a digital version of themselves, called an avatar. The avatar can be inserted into existing Shorts or used to generate new ones. YouTube says avatars will "look and sound like you" and presents them as a safer and more secure way to use AI.

AI researchers report that online content creators are fabricating people, including sexualized images of women in camouflage, and placing them in military contexts. The images draw large audiences and can earn money or act as propaganda, including idealized portrayals of political figures such as Donald Trump, even when viewers know they are not real.

Researchers propose a deepfake face detection method that recasts detection as a Visual Question Answering task for a Visual Language Model. Component-specific prompts direct attention to the eyes, nose and mouth, and a Q-Former module adjusts visual feature focus according to those prompts. The authors report that the method outperforms existing techniques in detection accuracy and robustness, though the source text gives no numerical results.

Baltimore became the first major U.S. city to sue xAI over deepfake sexual images generated by Grok. The complaint, filed March 24 in a circuit court, accuses xAI of violating city consumer protection laws through deceptive and unfair marketing of Grok and X as generally safe, citing a "put her in a bikini" trend and a post by Elon Musk that the city says endorsed nudifying edits of real people. The city seeks statutory penalties and injunctive relief.

Researchers propose a Two-Stage Defense Framework (TSDF) against deepfakes that aims to keep active defenses effective after attackers retrain their models. The framework uses dual-function adversarial perturbations that distort forged results and also poison the data preparation step of an attacker's retraining pipeline. The source reports that traditional interruption methods degrade sharply under adversarial retraining, while TSDF shows stronger dual defense capability.

This ACM Computing Surveys paper, published in Volume 58, Issue 10 (July 2026), examines fairness in deepfake detection across ethnicity, gender and age. The source text provided contains only bibliographic details and no findings, method or results.

FauForensics is a framework for detecting audio-visual deepfakes that uses facial action units, which quantify facial muscle activity, as forgery-resistant features. It computes frame-wise audio-visual similarities through a fusion module with learnable cross-modal queries, aligning lip-audio relationships to address feature heterogeneity. Tested on four public datasets, it reports state-of-the-art results with a 5.17% average cross-dataset improvement over existing methods.

This paper proposes a triple-branch network for deepfake detection that jointly learns spatial features and frequency features from the original image and from images reconstructed through different frequency channels. The authors add feature decoupling and fusion losses derived from mutual information theory to make the model focus on task-relevant features. The method reportedly achieves state-of-the-art performance across six large-scale benchmark datasets.

The paper proposes FoVB (Forgery-aware Audio-Visual Adaptation with Variational Bayes), a framework for multi-modal deepfake detection. It models audio-visual correlation as a Gaussian latent variable estimated via variational Bayes, using difference convolutions and a high-pass filter to extract forgery traces from both modalities, then factorizes the variable with an orthogonality constraint. The authors report that FoVB outperforms other state-of-the-art methods across various benchmarks.

GenAI has made social engineering and phishing faster and more convincing, enabling hyper-personalized messages, cloned executive voices and realistic video impersonations. Deepfake incidents have already caused financial loss and operational disruption for organizations, and real-time face and voice cloning makes scams on collaboration platforms harder to spot. Check Point Services has expanded its training portfolio in response.

YouTube is expanding its likeness detection technology, which identifies AI-generated deepfakes, to a pilot group of government officials, political candidates, and journalists. Pilot members can request removal of detected content they believe violates YouTube policy, and YouTube will evaluate each request against its privacy policy guidelines, protecting parody and political critique.

YouTube is expanding its likeness detection feature, which scans for AI-generated deepfakes of a person's face and works like Content ID, to a pilot group of journalists, government officials and political candidates starting Tuesday. The feature is already available to millions of content creators, and YouTube declined to say who is in the pilot group.

Meta's Oversight Board says the company's methods for identifying deepfakes are "not robust or comprehensive enough" for fast-spreading misinformation during armed conflicts such as the Iran war. The Board is calling on Meta to overhaul how it surfaces and labels AI-generated content across Facebook, Instagram, and Threads. The call follows an investigation into a fake AI video of alleged building damage in Israel shared on Meta's platforms last year.