Pangram CEO Warns AI-Generated Content Is Pushing the Internet Toward a Trust Crisis

AI-generated text and images are moving through everyday online systems, appearing in job applications, product reviews, insurance claims, newsletters, and social feeds. That expansion is making authenticity harder to judge for platforms, employers, businesses, and readers, Pangram CEO Max Spero said in a TechCrunch Equity discussion.

Spero framed the issue as a widening trust problem rather than a simple contest between human and machine writing. The concern echoes broader commentary about AI “slop,” a term used for low-value automated material that can crowd out useful information and make online spaces harder to navigate. Coverage from The Ringer and The New York Times has likewise focused on the growing need to identify, filter, or clean up machine-produced content as its volume rises.

AI detection doesn't reduce neatly to a “real or fake” decision. Writing often combines human ideas with automated editing, drafting, translation, or research, creating a boundary between AI-assisted and AI-generated work that tools must interpret rather than simply label.

That distinction matters because a false accusation can damage a writer, applicant, researcher, or publisher. Pangram's product is positioned as a detection system for institutions that need signals about how content was produced, while the discussion around AI detection has also raised questions about accuracy, transparency, and how results are presented. The podcast episode featuring Spero focused directly on how detection works, while the TechCrunch interview examined where platforms should draw the line between assistance and authorship.

Substack is an early testing ground for content-origin tools because newsletters mix professional publishing, personal commentary, and AI-assisted workflows. Pangram recently raised $9 million for its detection system and formed a partnership with Substack, TechCrunch reported, putting the technology inside a platform where judgments about authorship can affect writers' reputations and audiences.

The partnership has also triggered resistance. Substack writers described the new detection approach as a “witch hunt,” according to 404media, reflecting fears that imperfect classifications could turn editorial platforms into systems of suspicion. The dispute captures the central trade-off: readers want more context about how writing is produced, while creators want control over their work and protection from opaque automated judgments.

Pangram is part of a group of startups building services that sit between generated content and the people who consume it. Their customers include platforms, employers, insurers, and other organizations that need to assess material at scale, especially where fabricated or automated submissions create financial, legal, or reputational exposure.

The $9 million financing gives Pangram resources to expand its detection system and pursue platform partnerships. It also shows that content verification is becoming a commercial category alongside generation, moderation, and identity services. Commentary from Derek Thompson and The Ringer places that market in a wider debate over whether online publishers can preserve useful human expression while automated systems produce material faster than people can review it.

Readers are being asked to evaluate not only whether content is accurate, but also how it was made. That extra layer changes the role of labels, provenance signals, moderation systems, and editorial review. A detection score can help an organization prioritize material for inspection, but it doesn't establish that a passage is false or that a person violated a platform's rules.

The New York Times' focus on a “slop janitor” reflects the practical work required to manage unwanted AI material after it reaches online communities. The same pressure appears in social feeds, commercial reviews, hiring pipelines, and claims processing. Platforms that adopt detection tools will have to explain their standards, give users a way to challenge decisions, and distinguish evidence of automation from evidence of deception.

The central debate will shift from whether AI content exists to who gets to define acceptable use. Pangram's Spero argues that the internet is approaching a dangerous point because automated material can overwhelm the signals people use to judge credibility. The response will involve detection, but also disclosure rules, provenance systems, human review, and platform design.

Those measures won't settle the authorship question by themselves. Writers, publishers, employers, and readers need policies that recognize editing and collaboration instead of treating every use of an AI tool as equivalent. Pangram's partnership with Substack gives the issue an immediate public arena, while the broader discussion across technology, media, and culture shows that online trust is becoming a shared infrastructure problem.