NEW YORK--(BUSINESS WIRE)--Pangram, the AI research lab known for its industry-leading 1 in 10,000 false positive rate, is announcing two new products. Pangram 4, its latest text detection model, offers major improvements in distinguishing AI-assisted writing and mixed human-AI content, plus greater robustness against AI humanizer programs. The company also released a research preview of an AI image detection model which is 99.5% accurate in internal benchmarks. Pangram 4 is available to all customers today, and Pangram AI Image Detector Research Preview is open to the public.
The launches follow a $9M funding round led by Menlo Ventures, with participation from Haystack, ScOp Venture Capital, Script Capital, and Cadenza. Pangram is using this capital to push forward on two fronts: improving the accuracy of its core text detection, as well as expanding into new modalities, starting with images.
“Fully AI-generated content is everywhere now, and much of it is undisclosed,” said Max Spero, cofounder and CEO of Pangram. “Pangram exists to make authorship legible for publishers, teachers, journalists, and anyone who relies on the truthful and authentic communication of information.”
Yet, as familiarity with AI writing has grown, so has the demand for products that obfuscate the true origin of AI-generated writing. Pangram 4 is trained to counter these kinds of “humanizer” tools, accurately identifying AI-written text even after it has been passed through a tool that paraphrases, rewords, or otherwise modifies it in order to evade detection. The new model also delivers major accuracy gains on mixed human-AI writing, while still maintaining a 0.01% false positive rate.
Menlo Ventures partner Deedy Das frames the problem in terms of trust. "With LLMs, the cost of producing language collapses to zero, and the cost of trusting language goes to infinity," he said. "Pangram has emerged a clear #1 in quality and is trusted across Substack, Quora, and other platforms to power detection across their website."
Currently in research preview, Pangram’s AI Image Detector represents a new frontier for the company. “AI image detection is challenging because real and AI content is often blended, like in a photo of an AI-generated ad on a billboard,” said Bradley Emi, cofounder and CTO of Pangram. “That’s why we tried to cast as wide a net as possible: whether you're browsing social media or assessing photographic evidence in a courtroom, everyone deserves to know what's real.”
Pangram 4 is live today for all customers, and the research preview for Pangram’s AI Image Detector is open to the public for testing and feedback. As generative AI continues to proliferate across every format, Pangram plans to bring the same standard of reliability to detecting all modalities of AI-generated content. To try Pangram 4 or Pangram AI Image Detector Preview, visit www.pangram.com.
About Pangram
Pangram Labs is the technology leader in AI detection systems, surpassing other detection providers in accuracy and reliability. Pangram’s detection systems are used today by thousands of businesses. Founded by classmates at Stanford University, Pangram has been adopted by academic institutions, content moderators, journalists, media, and consumers concerned with transparency.
Also Read:
AI-Generated Code Security: What are the Risks, Challenges, and Solutions?
