Pangram, the leading AI text detection platform, has released an upgraded model that significantly enhances its already unmatched ability to identify AI-generated content. CEO Max Spero said the new system is even more accurate at spotting outputs from advanced models by OpenAI, Anthropic, and Gemini — including some not yet publicly available.
Spero stated that Pangram’s advantage lies in its use of “synthetic mirrors” instead of traditional metrics like perplexity, making it highly adaptable. The update also includes a new “mixed” classification in educator dashboards, helping teachers detect AI-human hybrid text with precise breakdowns. Spero said that those relying on AI and paraphrasers to evade detection won’t succeed anymore.
