AI's Napster Moment: What Music Piracy Teaches Investors About Anthropic's IPO

Anthropic's IPO filing contains a remarkable admission: the firm's own AI models may "pose catastrophic or existential risks to humanity." That's a startling line for a prospectus. But before investors get lost pricing Armageddon, they should focus on nearer-term risks that are easier to model and already visible in the data.

Two stand out. First, frontier AI is dismantling the intellectual property framework underpinning the entire software industry. Second, Anthropic itself, having played executioner to legacy software firms, may become the next victim. For both, the music industry's brutal disruption in the early 2000s offers a remarkably precise roadmap.

The disruption timeline is already well established. OpenAI's Operator and Anthropic's Claude Code, both released in early 2025, let non-specialists "vibe-code" using natural language prompts alone. By January 2026, Anthropic's Claude Cowork agent shipped with plug-ins capable of specialized legal, compliance, and sales work. The result was the "SaaSpocalypse," a sharp selloff across software-as-a-service stocks as investors recalculated what a subscription model was even worth.

The latest frontier models raise the stakes considerably. They pose what amounts to an existential threat to cybersecurity. Recent incidents have shown AI systems accelerating their ability to exploit code vulnerabilities at scale, raising fears the internet becomes what one commentator called a "terminally hostile environment" for cloud-connected devices. That's a genuine risk, though arguably a solvable one. The same AI capabilities that find the bugs can patch them, meaning the cybersecurity crisis should resolve itself over time, albeit expensively and messily.

The Decompilation Problem

Far harder to solve is AI's growing ability to decompile, reverse-engineer, and clone existing software outright. Microsoft Xbox CEO Asha Sharma put the commercial stakes bluntly in a "Code Red" staff memo circulated two weeks ago. "In recent weeks, we've seen games rapidly decompiled and rebuilt to run on hardware they were never originally designed to support," she wrote. Some of the industry's biggest titles, she noted, have already been "remixed into the source code of another."

That points to a scenario where proprietary code becomes open-source whether companies consent or not. Open-source software already underpins significant infrastructure. Most public cloud computing runs on variants of Linux. But the bulk of commercial software is sold as proprietary code, licensed or subscribed to rather than given away. If AI converts that code into what economists call a non-excludable good, nearly every software business model built over the past three decades gets upended.

This is where the music industry's experience becomes instructive. Starting in the late 1990s, digitization and streaming made recorded music effectively open-source overnight, reshaping every corner of the business. Stephen Witt's 2015 book "How Music Got Free" documents the mechanics well. German researchers invented the MP3 algorithm in 1995, collapsing the cost of digitizing and reproducing music. Napster launched in 1999, enabling peer-to-peer file sharing and triggering a piracy wave so severe it briefly looked like nobody would ever pay for recorded music again.

Copyright litigation eventually bankrupted Napster in June 2001. But the labels' legal win didn't restore the old order. Over the following decade, the industry restructured completely. Storied labels like EMI and PolyGram, built on captive artists and blockbuster album cycles, collapsed. Streaming platforms like Spotify and YouTube rose in their place.

Three lessons from that history apply directly to software investors weighing AI exposure today. First, it's genuinely hard to predict where sustainable value lands once disruption starts. Experienced venture capitalists and sophisticated corporate partners who bet on Napster's profitless, get-big-fast model lost their investments entirely. Apple's iTunes initially tried porting the old album-ownership model into digital form and hit a dead end. It took roughly a decade for subscription and ad-supported streaming to prove viable. No single strategy should be treated as a sure bet.

Second, consider the fate of the disruptor itself. Napster didn't survive its own disruption. It collapsed under copyright infringement lawsuits in 2001. Anthropic now faces an eerily similar wave of litigation. The settlements the company has absorbed so far over its use of copyrighted training data, including a $1.5 billion settlement approved by a US judge in July, shouldn't comfort investors. If courts eventually rule that cloning closed-source software itself violates intellectual property law, the financial exposure could be an order of magnitude larger than anything seen to date.

Third, and this is where Anthropic's position looks worse than Napster's ever did: streaming technology was a weapon a startup could point exclusively at incumbents. AI cuts both ways. Nothing stops rivals from turning the same tools on the frontier labs themselves. Chinese labs are already reverse-engineering US models and releasing open-source versions. MoonshotAI's Kimi K3.1 Pro, released last week, reportedly trails the most sophisticated US models by just five weeks in capability. MIT's Christian Catalini summed up the dynamic sharply: "the closed, excludable frontier that gets IPO investors excited keeps showing up as a free download."

The Bottom Line

Marc Andreessen's line that "software is eating the world" is now fifteen years old. The AI era adds a darker twist: the eaters are increasingly vulnerable to being eaten themselves. Investors weighing Anthropic's IPO need to price in more than catastrophic tail risk from superintelligent models. They need to ask whether Anthropic's own closed-source moat, the thing justifying its valuation, can survive the same forces it unleashed on everyone else. History from the music industry suggests the disruptor rarely gets to keep the spoils. Value migrates to platforms and models nobody has invented yet, and the companies that triggered the disruption are often first against the wall.