Pangram, a New York-based AI detection startup, has raised $9 million in funding led by Menlo Ventures, with participation from Haystack, ScOp, Script Capital, and Cadenza. The round coincides with the launch of Pangram 4, the company’s next-generation AI text detection model, alongside Pangram Image, a new tool for detecting AI-generated images.
Co-founder Max Spero said the updated text model exceeds 99% accuracy in identifying AI-assisted and mixed human-AI writing, while also improving detection of AI humanizer tools designed to evade detection. Founded roughly two years ago by Stanford graduates Spero and Bradley Emi, Pangram trains its models by comparing authentic human writing against AI-generated versions matched for topic and tone, allowing it to identify stylistic patterns without relying on metadata or watermarks.
Spero said the goal is not to discourage AI use but to ensure transparency, noting that disclosed AI assistance can be acceptable. He said the technology aims to preserve human-generated content as AI output continues proliferating across the internet.
Pangram’s image detection model, currently in research preview, identifies AI-generated visuals using pixel-level statistical analysis rather than watermark detection, working across models regardless of origin. The company’s technology is already integrated by clients including Substack and Quora, alongside universities, publishers, and recruiters, and competes with detection tools from Winston AI, Originality.ai, Copyleaks, and GPTZero.
Featured image: Credit: Pangram
