New York-based AI detection startup Pangram raised $9 million in July 2026 and launched two new detection models: Pangram 4, its next-generation AI text detector, and Pangram Image, an AI image detection model currently available in research preview.
The funding round was led by Menlo Ventures, with participation from Haystack, ScOp, Script Capital, and Cadenza. Pangram was founded roughly two years ago by Stanford AI and machine learning graduates Max Spero and Bradley Emi, following the release of ChatGPT.
Pangram 4 claims over 99% accuracy in identifying AI-assisted writing and mixed human-AI content, and is designed to detect AI humanizer programs that attempt to disguise machine-generated text. The company says roughly one in 10,000 human documents are incorrectly flagged as AI-generated. The image detection model works on pixel-level distributions to identify statistical differences between real photos and AI-generated images — and, according to Spero, can detect an AI-generated image embedded within a real-world photograph. A wider release of the image model is planned for the coming weeks.
Pangram’s detection system was trained on tens of millions of known human documents, with a “synthetic mirror” created for each — matching topic, length, and tone but written by a large language model — allowing the system to learn stylistic patterns characteristic of AI-generated content.
The service is available to consumers via a $20-per-month subscription or a Chrome extension that labels posts in real time on platforms including X, LinkedIn, Substack, Reddit, and Medium. Pangram also offers API access; current integrations include Substack, Quora, schools and universities, publishers, and recruiters.
The launch comes as institutions begin enforcing stricter rules around AI-generated content. Academic preprint server arXiv introduced a policy in 2026 that can impose a one-year submission ban on authors whose work shows signs of unreviewed LLM output. Spero framed the stakes broadly: “If we do not actively discriminate in favor of human content, then we’re just gonna see more and more AI, and it’s just gonna drown out any human signal that we have.”
Source: TechCrunch