AI Detection Startup Pangram Secures $9M to Spot Machine-Written Text and Visuals

New Funding, New Models

A New York startup called Pangram wants to help you tell the difference between a human-written article and something a chatbot churned out. Menlo Ventures led the funding round, joined by Haystack, ScOp, Script Capital, and Cadenza as additional backers. The company started in 2024, shortly after ChatGPT debuted.

The timing makes sense. AI-generated content has exploded across the internet. Low-quality SEO articles, bot posts, and even disinformation campaigns that use large language models are everywhere. Max Spero, one of Pangram's co-founders, said, "LLM-powered Russian disinformation campaigns and UAE-influenced campaigns on Twitter" are just one example of what detection tools need to catch.

Competitors such as Winston AI, Originality.ai, Copyleaks, and GPTZero offer similar services, but Pangram differentiates itself with its low false-positive rate and backing from prominent venture firms.

How Pangram's Detection Works

The new text detection model looks for the subtle patterns that AI writing tools tend to repeat. "Our model is learning the stylistic differences and the choices that AI makes consistently," Spero said. That lets it flag AI-generated text with high confidence.

Get the market news that matters in a five-minute read with Market Briefs, our free daily newsletter

It also handles the gray area where humans and AI mix. In one test, a human-written article that had been polished by an AI chatbot scored just 13% on Pangram's AI-assisted scale. The same article in its original, unpolished form scored 100% human.

False positives are a big worry with these tools. Nobody wants their own work wrongly flagged as machine-made. The service costs $20 per month for users who want access through the web or a browser extension.

Who Is Already Using It

Pangram is not starting from scratch. Several well-known platforms have already plugged in.

Quora relies on the API to screen content. Additional API clients include educational institutions, book and article publishers, literary agents, and hiring managers.

Spero put it plainly: "I think it's just incredibly valuable to know whether what you're looking at is something that's AI-generated or not." He added that knowing the source changes how you read something - whether you need to watch for hallucinations or can trust it as well-researched reporting from an actual journalist.

What This Means for the Content You Read

Pangram plans to release its image detection model more broadly in the coming weeks. The image preview is already out, and it will go wider soon.

The bigger picture: AI content is not going away. Spero expects it to keep proliferating. "We're getting new GPUs faster than new people are being born," he said. "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."

Join Market Briefs, our free daily newsletter, for a quick daily rundown of the markets