Pangram launches AI video and image detection models

What launched? AI-detection company Pangram is officially launching two new models that detect AI-generated imagery and video, moving beyond the text detection it is best known for. The flagship image model, Image 1.0, is designed to handle the degraded quality of AI images as they circulate online and scored 99.96% on an internal benchmark, according to CEO Max Spero. A video model, Image Flash 1.0, works by sampling frames from a video and scored 98.57% on one internal benchmark.

Why move beyond text? Spero, who co-founded Pangram with CTO Bradley Emi, told Pirate Wires the company began working on image detection at the start of 2026 because AI images were becoming increasingly hard to spot, and because people are using them to create hoaxes and fake war footage. He also pointed to e-commerce platforms such as Amazon, where ads increasingly use AI-generated images. When Pangram benchmarked existing image detectors, Spero said most reached only 80 to 95% accuracy.

Context Image and video detectors from rivals such as TruthScan have been on the market since 2025 or earlier; Spero said Pangram deliberately waits to release only what it considers state-of-the-art by a large margin. In July 2026 Pangram released a research preview of Pangram Image alongside its Pangram 4 text model and announced a $9M round led by Menlo Ventures. Pangram Image analyses images pixel by pixel rather than relying on watermarks or metadata, shows a heat map of AI-flagged regions, and covers generators including OpenAI's GPT Image, Google's Nano Banana, Midjourney, FLUX and Grok Imagine, as well as frames from video models such as Kling, Seedance, Veo and Wan.

Sources Pirate Wires (The Industry): Exclusive: Pangram Can Now Detect AI Video
Pangram: Introducing Pangram Image Detection (Research Preview)
Pangram AI Image Detector