The model achieves 95.2% accuracy, outperforming global competitors across benchmark metrics
StealCut, a 2026 SelectUSA Pitch Competition-winning AI security startup, released its new AI image detection model, achieving 95.2% accuracy and outperforming three leading global detection services in March 2026 benchmark testing.
The benchmark tested AI-generated images from thirteen of the latest generative AI models, including Grok, Gemini, and ChatGPT, as well as outputs from commercial deepfake websites. According to the results, StealCut outperformed the three global detection services across all measured categories, including overall accuracy, forged-image detection rate, and average detection performance across generative AI models.
The launch follows StealCut’s first-place win in the global pitching competition at the SelectUSA Investment Summit 2026, hosted by the U.S. Department of Commerce, as the company accelerates its entry into the U.S. market. StealCut’s technology is also being deployed in a project with SBS, one of South Korea’s largest broadcasters.
In the June 2026 benchmark, StealCut’s model recorded 95.2% overall accuracy and a 91.7% detection rate for forged images, outperforming the second-ranked service by 7.7 percentage points. The model also achieved a 96.5% average detection rate across the latest 13 generative AI models.
Also Read: AiThority Interview with Matej Bukovinski, Chief Technology Officer at Nutrient
As the latest generative models produce images that are increasingly difficult for humans to distinguish from authentic content, StealCut is designed to detect AI-generated images at the point where manual review becomes unreliable. Rather than optimizing only for fixed test sets, the company continuously benchmarks its model against newly released generative tools and commercial deepfake sites, prioritizing performance in real-world production environments.
StealCut also holds proactive deepfake defense technology protected by five patents filed in South Korea, the United States, and the European Union.
The rise of AI-generated and AI-edited images has created new risks for newsrooms, online platforms, customer support teams, marketplaces, and content review operations. False or manipulated images can contribute to misreporting, fraud, abuse, and reputational damage before they are identified.
StealCut’s detection tool is designed to serve as a lightweight verification layer within existing workflows. It can help editorial teams, platforms, and businesses flag manipulated images in abuse reports, listings, claims, user submissions, and other digital content before they cause harm.
The technology also supports companies building image and video databases. As synthetic content becomes more prevalent online, identifying AI-generated content is increasingly important for maintaining training data quality and protecting the performance of AI models built on that data.
StealCut’s detection tool is available now at www.stealcut.com, where users can upload an image and receive instant results.
Also Read: AI systems – Interoperable AI systems: Connecting models across platforms
[To share your insights with us, please write to psen@itechseries.com ]
Comments are closed.