Pangram

Pangram is a technology company that provides software to detect artificial intelligence-generated content and plagiarism. Founded in 2023 by former machine learning researchers from Google and Tesla, the platform helps educational institutions, publishers, and enterprises verify the authenticity of text and images. By analyzing structural, stylistic, and semantic patterns, the software accurately identifies whether content was written by a human or an artificial intelligence. The tool provides a detailed analysis that highlights specific segments of text that received artificial intelligence assistance rather than just providing a simple binary score, giving users a complete picture of document authenticity. The core classifier utilizes a traditional language model architecture trained on an initial diverse dataset of 1 million publicly licensed human-written documents, alongside artificial intelligence outputs. This neural network can detect text generated by popular frontier models and is actively benchmarked against 26 distinct language models, maintaining an accuracy rate exceeding 99 percent. The system is designed to identify text that has been processed by tools attempting to evade detection, often referred to as humanized text. Alongside written content analysis, Pangram offers an application to identify AI-generated images and includes a comprehensive plagiarism checker for simultaneous verification of original sources. The software operates across more than 20 distinct languages, establishing it as a globally applicable solution for a variety of demanding industries. Its primary user base includes academic institutions looking to maintain fairness in assessments, as well as media organizations striving for content transparency and editorial integrity. Beyond the classroom and newsroom, the platform provides specialized detection capabilities tailored for law firms, human resources departments, and compliance teams. To further support academic environments, Pangram provides resources for universities looking to establish institutional policies regarding appropriate technology use. Verification by independent academic researchers at institutions like the University of Chicago and the University of Maryland confirms the platform's reliability and its notably low false positive rates. Users can access the platform through a web dashboard, directly within learning management systems, or via programmatic interfaces. The learning management system integrations seamlessly connect with popular academic platforms such as Canvas, Moodle, Google Classroom, and Brightspace. For organizations processing high volumes of text and documents, Pangram provides a dedicated developer interface with flexible scaling options. Individuals and content moderators can also deploy the Browser Extension for Chrome and Firefox, sometimes referred to as Feed Scanner, which automatically labels artificial intelligence content on websites like X, LinkedIn, Substack, Reddit, and Google Docs. Based in Brooklyn, New York, Pangram operates with a core mission of mitigating the negative impacts of generative artificial intelligence spam while increasing overall content transparency on the internet. The company is driven by a team of seasoned machine learning engineers who continuously retrain their detection models with active learning techniques and hard negative mining. This ensures the detector avoids relying on basic perplexity metrics, which often fail in practice. Through continuous iterative updates, the software keeps pace with the latest frontier model releases, transparently publishing all benchmarking performance metrics on a public model card.