📰 Key Highlights
Pangram is a New York startup focused on detecting AI-generated content. It recently closed a $9M funding round led by Menlo Ventures, with participation from Haystack, ScOp, Script Capital, and Cadenza. Alongside the raise, Pangram launched its next-gen text detection model Pangram 4 and image detection model Pangram Image. According to Pangram, the new models achieve over 99% accuracy in detecting “AI-assisted writing” and “human-AI hybrid content,” and can more easily see through the disguises of AI humanizer tools. The image detection model is currently available only as a research preview, with wider release planned in the coming weeks.
Pangram was founded roughly two years ago by Stanford AI and machine learning grads Max Spero and Bradley Emi. The trigger: after ChatGPT launched, the internet was flooded with bot accounts, AI-generated SEO spam, and what Spero calls “LLM-driven Russian disinformation campaigns and UAE-backed Twitter influence operations.” At the core of their tech is a large ML model trained on tens of millions of known human-written documents. For each one, they build a “synthetic mirror” — same topic, length, and tone, but rewritten by top-tier LLMs. This way the model learns the consistent stylistic and word-choice patterns of AI, rather than relying on copy-pasted metadata or hidden watermarks. Spero emphasizes that the goal isn’t just to flag whether an entire piece is AI-written, but also to gauge the "