Grok AI Fuels 87% of Deepfakes in Early 2026

In the first half of 2026, Elon Musk's Grok AI was identified as the source for 87% of AI-generated files linked to deepfake attacks. A report by Resemble AI examined 1,760 news reports, uncovering 821 separate attacks affecting over 15,736 victims and generating approximately 3.46 million synthetic media files. This figure pertains specifically to files that researchers could count and attribute to a particular tool.

Concerns are amplified by the fact that one in six attacks, totaling 137 incidents, involved non-consensual sexual imagery, including that of children. Grok has previously faced significant criticism for generating explicit content upon user request. Regulatory bodies, including the European Commission and California Attorney General, have launched investigations into the platform's role in creating and spreading such material.

Resemble AI estimates potential civil liability for companies involved could reach $2.24 billion under US law, though verified direct financial losses were noted at $6.95 million. Many incidents resulted in reputational damage and exploitation rather than direct financial scams.

The reach of deepfake attack coverage nearly matched the entire previous year's total, with political disinformation incidents achieving a massive media reach. Corporate deepfake fraud, however, has largely disappeared from public reporting, possibly due to companies' reluctance to disclose successful attacks, suggesting such fraud may be occurring covertly.

Alongside the report, Resemble AI launched DETECT-World, a new system designed to identify AI-generated content, even from novel generators, by analyzing physical scene consistency and lighting. While internal benchmarks show high accuracy, external validation is pending.