Pangram banks $9M as new model claims to catch 99.66% of AI text

Pangram has raised $9 million to scale its AI detection software, and alongside the funding it released Pangram 4, a new AI text detection model, plus an AI image detection model in research preview, TechCrunch and commons.now both reported. Superintelligence News added that the raise comes as synthetic content floods the web and demand for detection surges. According to machinebrief.com, the company claims Pangram 4 detects 99.66 percent of AI-generated text with just one false positive per 24,000 documents. That is a striking accuracy claim for a category where false accusations of AI authorship carry real consequences. "Pangram 4 detects 99.66 percent of AI-generated text with just one false positive per 24,000 documents, the company claims." The precision figure is central to Pangram's pitch, since more sensitive detectors often flag human writing by mistake. machinebrief.com also reported that the model resists "humanizer" tools designed to disguise AI writing as human, a growing arms race as such tools spread. "The model also resists 'humanizer' tools that disguise AI writing as human." That positions Pangram 4 as an answer to writers who try to launder AI text through rewriting services, though the claims are the company's own. Along with the new models, machinebrief.com said API prices will rise two- to tenfold, reflecting a premium for the more capable detection. TechCrunch and commons.now, meanwhile, framed the raise as a bet on a market where AI content keeps flooding the internet, and Superintelligence News said the new tools are aimed at spotting both AI-written text and images.

4 sources•1 AugRead the Baker's Take→