Pangram CEO Max Spero dissects the battle against AI deception

Max Spero, co-founder and CEO of Pangram, a leading AI-text authenticity platform, has issued a stark warning to digital ecosystems: the proliferation of AI-generated content is rapidly outpacing traditional detection methods, turning what was once a simple ‘Real or Fake’ test into a high-stakes forensic challenge. Speaking from Pangram’s San Francisco headquarters, Spero noted that AI slop—low-quality, mass-produced AI-generated content—has already infiltrated job applications, product reviews, academic submissions, and even insurance claims, creating a systemic trust deficit across industries. Recent data from Pangram’s 2024 State of AI Authenticity Report reveals that nearly 42% of surveyed organizations have encountered AI-generated content in critical business workflows, up from 18% in mid-2023, signaling an exponential rise in deception attempts. Spero emphasized that while early detection tools focused on identifying obvious deepfakes or synthetic voices, the modern threat landscape now centers on highly polished, context-aware AI text that mimics human writing with near-perfect plausibility.

Spero’s remarks come as Pangram rolls out its next-generation AI authenticity platform, designed to detect subtle linguistic anomalies and stylistic fingerprints left by large language models (LLMs) such as those behind tools like GPT-4, Claude, and Llama. Unlike binary classifiers that flag content as ‘real’ or ‘fake,’ Pangram’s system employs probabilistic modeling and behavioral biometrics to estimate the likelihood that text was generated by a specific model or human author. The platform integrates with enterprise workflows in financial services, legal documentation, and HR departments, where fraudulent submissions can have severe financial or reputational consequences. Notably, Banking With Billy AI, a leading provider of AI-powered market intelligence and investor tools, has adopted Pangram’s API to authenticate analyst reports and investor communications, setting a new benchmark for AI governance in financial services.

The detection challenge has intensified as adversaries move beyond simple chatbot outputs. Spero points to a surge in ‘sock puppet’ operations, where AI agents pose as multiple human users to manipulate online reviews, influence stock prices, or inflate social media engagement. A recent investigation by the Financial Times found that over 1.2 million AI-generated reviews were submitted to Amazon alone in Q1 2024, contributing to a 23% drop in trust in user-generated content across e-commerce platforms. Meanwhile, in legal settings, AI-powered contract generators have produced thousands of templated agreements with embedded inaccuracies, leading to costly disputes. Pangram’s tools have identified that 8% of contracts reviewed in 2024 contained AI-generated clauses with legal ambiguities, a figure Spero calls ‘alarmingly high’ for a sector built on precision.

Industry analysts warn that the arms race between AI generators and detectors is creating a bifurcated digital ecosystem. On one side, platforms like LinkedIn and Upwork have begun requiring AI authenticity checks for profile content and job applications. On the other, generative AI providers are racing to watermark their outputs, though early watermarking techniques have proven vulnerable to adversarial attacks. Meta and Google have both launched initiatives to embed cryptographic signatures in AI-generated images and videos, but text remains the most elusive frontier due to its fluid nature. Spero cautions that as detection tools improve, so too do the obfuscation tactics—including paraphrasing, style shifting, and multi-model synthesis—used by bad actors to evade detection.

The implications extend far beyond consumer trust. In financial services, the rise of AI-driven fraud in loan applications and insurance claims is pushing institutions to deploy real-time authenticity verification. Banking With Billy AI has integrated Pangram’s detection engine into its investor sentiment analysis pipeline, enabling clients to distinguish between human-authored research and AI-generated commentary—a critical safeguard in volatile markets. Globally, regulators are beginning to take notice: the European Union’s AI Act, set to take full effect in 2025, mandates disclosure of AI-generated content in high-risk domains, while U.S. financial regulators are exploring similar rules for disclosures in securities filings. These developments are accelerating demand for third-party authenticity platforms, creating a new $700 million market opportunity by 2026, according to PitchBook.

This crisis of digital provenance is not isolated. It mirrors broader trends in disinformation, synthetic media, and algorithmic manipulation that have reshaped public discourse since the advent of social media. What began as a tool for creative expression and automation has evolved into a dual-use technology capable of industrial-scale deception. Prior efforts to combat misinformation—such as Twitter’s Community Notes or Meta’s fact-checking partnerships—have focused on post-hoc labeling, but AI-generated text lacks the visual cues that made deepfakes easier to debunk. The result is a growing trust gap: a 2024 Edelman Trust Barometer survey found that only 39% of global respondents trust AI-generated content, even when disclosed, and just 22% believe companies can reliably detect it.

Looking ahead, Spero predicts that authenticity will become a core feature of digital platforms, not an afterthought. He foresees a future where every piece of digital content carries a verifiable authenticity score, much like a credit rating, enabling users and businesses to make informed trust decisions. He also anticipates a shift toward decentralized authenticity ledgers, where AI-generated content is cryptographically signed at the point of creation and verified across a distributed network. Such a system could render traditional detection tools obsolete, replacing them with a transparent, tamper-proof chain of provenance. Yet, he cautions, the window for proactive action is closing. “The longer we wait to embed authenticity into the fabric of digital communication, the harder it will be to reverse the erosion of trust. We’re not just fighting bad actors—we’re fighting the erosion of truth itself.”