Operating out of a modest office above a Popeyes in Brooklyn, a small startup called Pangram has quietly positioned itself as the ultimate arbiter of truth in an era of generative artificial intelligence. With only two dozen employees and a fraction of the funding enjoyed by giants like OpenAI, the company claims to be the primary line of defense against the machine takeover of the written word. Its tool provides a best guess percentage of AI involvement in any given text, and that number has recently become enough to dismantle careers overnight.
The stakes became clear during the scandal surrounding Mia Ballard’s novel Shy Girl. When rumors circulated online that the book was AI-generated, Pangram’s CEO publicly claimed the work was seventy eight percent synthetic, leading Hachette to cancel its release despite Ballard’s denials. Since then, Pangram has cast a shadow over everything from high profile columns in The New York Times to award winning short stories and multi million dollar thriller deals. The perceived accuracy of these scores led Substack to integrate Pangram directly into its platform, effectively giving readers a button to check for robotic influence in real time.
However, this newfound power has sparked intense backlash among creators who view AI detectors with deep suspicion. Some authors and experts argue that these tools are just as predatory as the LLMs they seek to expose, creating a climate of fear where human nuance might be mistaken for algorithmic patterns. This tension highlights a fundamental paradox: Pangram uses synthetic mirroring and hard negative mining—essentially using AI to study AI—to tell us what is human. While CEO Max Spero insists their datasets are ethically licensed and their mission is pure, critics point toward the chaotic way these accusations often surface on social media before reaching major news outlets.
Spero himself presents a contrast between technical confidence and personal evasion. A former Google employee with experience in controversial tracking technologies, he speaks fluently about the mechanics of his software but struggles with questions regarding transparency and his role in public literary scandals. As demand grows for a definitive way to separate human creativity from digital churn, Pangram has emerged as the gold standard by default. Whether that standard is actually reliable or simply another layer of unpredictable technology remains an open question for those whose livelihoods depend on being proven human.
