
Pangram detects AI-written text with percentage scores.
It has sparked scandals in publishing, including canceled deals.
Critics worry about false positives and bias.
What happened
Pangram, a 24-person AI detection startup based above a Popeyes in Brooklyn, has raised $13 million and emerged as a front-runner in identifying AI-written text, scoring high-profile works like a Hachette novel (78 percent) and a New York Times column (100 percent).
Why it matters
Publishers, agents, and platforms like Substack are integrating Pangram, and its scores have led to canceled book deals and public accusations, raising concerns about false positives and bias against non-native and neurodiverse writers, as well as writers of color.
What to watch
Pangram's CEO Max Spero offers a cash reward to anyone who can prove their flagged work is original, but so far no one has claimed it, while the company plans to expand its workforce by 25 percent with six new job openings.
Ask the AI about this article →
Pangram's rise reflects a growing demand to distinguish human from AI writing, but its influence has ignited fierce debate. The startup, founded by Max Spero and Bradley Emi after ChatGPT's launch, has become a gatekeeper in publishing, with agents like Todd Shuster using it to screen manuscripts. However, its scores have led to canceled deals and public accusations, raising questions about accuracy and fairness.
The controversy centers on Pangram's role in high-profile cases, from Mia Ballard's 'Shy Girl' to a $2.4 million thriller deal. While Spero downplays its impact, critics argue that false positives can destroy careers, and studies suggest bias against certain groups. The company's integration with Substack and partnerships with literary agencies could expand its reach, but the lack of transparency and potential for error remain significant concerns.
As AI becomes more prevalent in writing, Pangram's technology may become more necessary, but its reliability and ethical implications are far from settled. The company offers a bounty for false positives, yet no one has claimed it, which could signal confidence or a chilling effect. Whether Pangram's judgment can be trusted without independent oversight remains an open question.
For example, today's edition would include:
AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.
Free · takes 30 seconds · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. Q&As are published on this page for other readers too.
At the Semicon Network Summit ahead of SEMICON Taiwan 2026, President Ching-Te Lai said Nvidia invests and pro…

Gradiant announced a new contract to deliver a full turnkey water and wastewater package for a hyperscale AI d…

SK Hynix holds ~55% of the High Bandwidth Memory (HBM) market and is the leading memory supplier for AI hardwa…

Saudi Arabia's LEAP tech conference opened with $15 billion in planned technology investments

NTT group has started offering a cloud service for physical AI, aimed at supporting the development of robots…

Pasqal, four days after listing on Nasdaq, came to SEMICON Taiwan to argue that the industry's assumption that…
