A new Resemble AI report puts Elon Musk's Grok at the center of publicly documented deepfake misuse in early 2026, according to Tom's Guide. The report claims Grok generated 87% of traceable AI-created files connected to documented deepfake attacks during the first half of the year. The scope is important. Tom's Guide says Resemble AI examined 1,760 news reports and identified 821 separate attacks involving at least 15,736 documented victims. Those incidents included roughly 3.46 million synthetic images, videos and audio files. The 87% figure applies only to files researchers could both count and attribute to a particular tool, and the dataset is based on publicly reported incidents rather than a complete accounting of all deepfake activity. Even with that limitation, the numbers describe a misuse pattern around widely available AI media generation. Resemble AI found that 137 of the 821 attacks — about one in six — involved non-consensual sexual imagery of adults or children, according to Tom's Guide. The report also says many incidents were not straightforward financial scams, but cases involving reputational harm, harassment or sexual exploitation. Tom's Guide reports that Grok had already faced criticism after users asked the chatbot to digitally undress people or place them in sexualized scenarios. Some images reportedly depicted minors. The outlet says that prompted scrutiny from regulators in multiple jurisdictions. That scrutiny includes a formal European Commission investigation into X over sexualized images generated with Grok, according to Tom's Guide. The outlet also reports that California Attorney General Rob Bonta opened an investigation into the creation and spread of non-consensual explicit material through the platform, with authorities in the UK, Australia and other countries also scrutinizing the tool. The financial exposure described in the report is mostly potential rather than realized. Tom's Guide says Resemble AI estimated that companies permitting or distributing the documented images could face as much as $2.24 billion in potential civil liability under US law. Verified direct financial losses identified by researchers were much smaller, totaling $6.95 million. The report also separates attention from direct losses. Resemble AI calculated that coverage of the attacks had a potential reach of 292.8 billion during the first half of 2026, nearly matching the 296.4 billion recorded across all of 2025, according to Tom's Guide. Political deepfakes were among the most widely covered: the report identified 22 political disinformation incidents, each with potential media reach above 1 billion people. Corporate deepfake fraud, by contrast, appeared less visible in public reporting. Tom's Guide says Resemble AI found 15 publicly reported corporate incidents, with 14 appearing in only one publication. That does not prove corporate fraud is disappearing; within the provided evidence, it only shows that such incidents were rarely visible in public news coverage in the dataset. Tom's Guide also frames the findings against Musk's larger financial bet on Grok. Citing Business Insider, it reports that after SpaceX's acquisition of xAI, Musk told employees that SpaceX's AI revenue could surpass revenue from all other products as soon as September, and that SpaceX reported $2.56 billion in AI revenue in its most recent quarter. Those business claims are separately reported in the item and are not independently corroborated in this cluster. Who benefits: Regulators, safety vendors and platform risk teams benefit from more public incident data to guide enforcement and controls. Resemble AI also gains visibility as the source of the dataset, though the figures remain single-source in this cluster. Who's exposed: Companies that permit or distribute non-consensual synthetic imagery are the most directly exposed under the report's liability framing. Platforms offering consumer-facing image generation also face scrutiny if safeguards fail to prevent sexualized or abusive outputs.