Pangram co-founder and CEO Max Spero used a September 2, 2026 podcast appearance to argue that AI detection has outgrown its original binary framing, and that the harder question — how much of a piece of writing was machine-generated versus machine-assisted — is where the market is actually heading. The startup recently closed $9 million in funding, signed a partnership with Substack to flag AI-written newsletters, and rolled out a new AI image detection tool, all within the past several months.
Spero's pitch is that a simple "real or fake" verdict is no longer useful. A human writer who runs a draft through a grammar model is not the same as a fully synthetic op-ed, and treating them identically breaks trust with the users being flagged. Pangram is positioning itself to grade shades of authorship rather than deliver a yes-or-no ruling — a shift that changes what detection tools compete on.
“The internet has a trust problem, and it's not just because social media feeds are filling up with AI slop.”— Max Spero, Pangram co-founder and CEO
The trust problem Spero describes is spreading well beyond social feeds. AI-generated text and images are landing in job applications, product reviews, and insurance claims, forcing platforms to decide what disclosure looks like when the writing is 30% machine versus 90% machine. Pangram is one of a small group of startups that emerged in the past couple of years selling themselves as the trust layer the open web now needs.
Key facts
- 01Pangram raised $9 million for its AI detection system, its largest funding milestone to date.
- 02Substack is now using Pangram's technology to flag which newsletter authors use AI to write.
- 03Pangram recently launched an AI image detection tool alongside its text classifier.
- 04CEO Max Spero appeared on TechCrunch's Equity podcast on September 2, 2026 to discuss the assisted-vs-generated distinction.
The Substack deal is the most public test of that thesis. Readers on the newsletter platform can now see indicators of which of their favorite authors are leaning on AI to draft posts, a disclosure layer that Substack chose to outsource rather than build in-house. That is a meaningful vote of confidence in Pangram's classifier at a moment when detection tools have taken repeated public hits over false positives against human writers.
The $9 million round funds the harder engineering problem Spero is now describing. Distinguishing "AI-assisted" from "AI-generated" requires not just a stronger classifier but a calibrated confidence score, a policy layer, and a UI that platforms can actually deploy without accusing their own users of fraud. It is a different product from the one-shot detectors that flooded the market in 2022 after ChatGPT's launch.
The image side is earlier. Pangram's recently released AI image detection tool enters a category where model outputs are moving faster than classifiers can keep up, and where watermarking standards from the major model providers remain fragmented. Spero has not disclosed accuracy benchmarks for the image tool, and the company will need third-party evaluation before enterprise buyers commit at scale.
Skeptics of the detection category note that every classifier eventually loses ground to the next model generation, and that adversarial prompting — asking a model to write in a deliberately human style — remains an unsolved problem across the industry. Spero's framing partially sidesteps that critique: if the product's job is to estimate the degree of AI involvement rather than to catch cheaters, the bar for usefulness is lower than perfect classification. But it also complicates the sales pitch to platforms that want a clean policy trigger.
The bet Pangram is placing is that trust infrastructure for the AI-generated web will be a durable line item, not a novelty purchase. If Substack's implementation drives measurable engagement or subscription lift, other publishing platforms and marketplaces will follow. If it triggers a wave of author complaints over misclassification, the entire detection category takes another credibility hit. The $9 million buys Pangram roughly one product cycle to prove which of those futures it is building for.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
