Pangram AI Detector Faces Trust Questions Amid Publishing Scrutiny
A Brooklyn-based AI startup, Pangram, is gaining traction for its AI detection technology. However, questions linger about the accuracy and ethical implications of its tools as the publishing world grapples with AI-generated content.
BROOKLYN, NY – Pangram, a burgeoning AI startup headquartered in Brooklyn, has rapidly positioned itself as a key player in identifying AI-generated text. With a lean team of 24 employees and $13 million in funding, the company, which was virtually unknown until early 2026, claims its technology can accurately assess the percentage of AI involvement in any given written piece. This capability has thrust Pangram into the spotlight, particularly within the publishing industry, where concerns about machine-generated content are escalating.
The company's visibility surged earlier this year following speculation on platforms like Reddit and YouTube regarding the novel "Shy Girl" by Mia Ballard. Pangram's CEO, Max Spero, asserted on X that the book was 78 percent AI-generated, a claim that led to Hachette canceling its traditional publication deal, despite Ballard's denials. Similar accusations soon followed, impacting publications like The New York Times, which was reportedly criticized for publishing an AI-assisted Modern Love column, with Pangram scoring it 100 percent AI-generated. Other high-profile texts, including a Commonwealth Short Story Prize winner (100 percent AI) and the novel "Daggermouth" (60 percent AI), also faced scrutiny.
In late July, Substack announced its integration of Pangram's detection tool, aiming to provide readers with a quick way to gauge potential AI usage in content. This move underscores the growing demand for such technologies, but it also raises significant debate among creators. "There is such distaste and anger at the AI detection software," commented Jane Friedman, an author and publishing expert. "There’s this feeling like they are just as evil, if not more evil, than the AI companies themselves." Friedman's sentiment reflects a broader unease about the authority and potential misuse of AI detection tools.
Navigating the Ethical Landscape of AI Detection
Max Spero, Pangram's 30-year-old cofounder and CEO, acknowledges the growing demand but also the inherent skepticism surrounding his company's work. In an interview, Spero explained that Pangram employs a technique called "synthetic mirroring," where large language models (LLMs) generate text closely matching human writing. This process trains Pangram's model on the nuances of AI-generated prose. Additionally, the company utilizes "hard negative mining" to identify and learn from false positives, refining its accuracy.
"All of our datasets are properly licensed, which I think is kind of rare in the AI world today," Spero stated, differentiating Pangram from some other AI firms. He also noted that Pangram's product is less data-intensive than major LLMs like ChatGPT or Claude. Spero, who previously worked at Google on the controversial FLoC initiative and later at the autonomous vehicle company Nuro, co-founded Pangram (initially Checkfor.ai) with Bradley Emi after identifying a business opportunity in the burgeoning AI content landscape post-ChatGPT's 2022 launch. They entered a crowded market, competing with companies like Originality.ai and GPTZero, but quickly emerged as a front-runner due to promising early testing.
Despite Pangram's growing influence, its involvement in high-profile cases has not been without controversy. Spero's initial public assessment of "Shy Girl" reportedly stemmed from a pirated manuscript, a detail he acknowledged but downplayed. When discussing the impact of Pangram scores on book deals, Spero suggested they are often a small factor among many. "If you talk to anyone involved, it is only a small, small part of the bigger picture," he said.
The integration with Substack highlights Pangram's strategy to embed its tools across content platforms. A Substack representative emphasized the platform's commitment to transparency, stating, "We simply believe you should know what you’re consuming." However, for many authors, the rapid advancement of AI detection technology raises existential concerns. A poll of 1,481 working writers conducted last year by Gotham Ghostwriters revealed that 61 percent use AI tools, with 7 percent having already published AI-generated text. This widespread adoption underscores the complex challenge facing industries like publishing, which are still adapting to the pervasive influence of artificial intelligence in creative processes.