The AI slop era has an official referee. Pangram released two new detection models this week, Pangram Image 1.0 and Pangram Image Flash 1.0, both built to identify AI-generated images and video. The flagship model is claiming 99.96% accuracy at detecting edits, compression, and fakes, while the smaller Flash model offers frame-by-frame video analysis at slightly lower accuracy.
Why does this hit so close to home? Because fake influencers, stolen comedy bits, and distorted war footage out of Iran and Ukraine are already floating through your timeline. Pangram’s CEO Max Spero told Pirate Wires that the shrinking margin for error forces constant improvement, and the company deliberately paced its releases so each model was genuinely state of the art. He noted the difference between 98% and 99.9% accuracy is enormous in this space.
For anyone who’s ever shared a clip and later found out it was fake, this is a welcome development. A detection tool gives you a way to check before you post, comment, or quote-tweet. Would you actually use a tool like this before sharing something that looked too crazy to be true?
About the Author
Local Lawton
Local Lawton is a contributor to LocalBeat, covering local news and community stories.