What’s actually converting unauthorised use into revenue is licensing markets, not myths and not slogans – and that’s true whether the unauthorised use is a torrented album or a scraped backlist.
An Iron Maiden story resurfaced in LinkedIn publishing industry circles this week, and it was such a warm and cozy story I almost wanted it to be true, despite knowing it was not.
First, and full disclosure, I have never knowingly heard Iron Maiden and have no opinion on their music. I’ve never read nor seen the video adaptations of Game of Thrones, which also crops up in this essay.
Iron Maiden, so the story goes, looked at where their music was being pirated most heavily, identified Brazil as a hotspot, and rather than sending lawyers, sent themselves – booking stadium tours across South America and making a fortune. Piracy, the moral runs, isn’t theft. It’s a map of unmet demand, and the smart move is to follow it.
It’s a wonderful story. It is also, in its most commonly repeated form as surfaced this past week, false.
What actually happened
In November 2013, Musicmetric – a UK firm that tracked BitTorrent and social-media activity – was quoted in The Guardian noting that Iron Maiden’s heavy file-sharing traffic in Brazil correlated with the band’s long-running touring success there. That’s a retrospective observation: the band had been touring South America intensively for years; the piracy data simply illustrated why.
Weeks later, the tech site CiteWorld ran a follow-up claiming the band had used that analytics data to plan the tours – a causal claim Musicmetric never made and has explicitly disowned. Musicmetric’s own head of PR put it on record: the company never worked directly with Iron Maiden, and the CiteWorld story was “sadly not substantiated.”
CiteWorld, TechCrunch and Rolling Stone all issued corrections. The retraction has been sitting in plain sight, on the record, since January 2014. But hey, why let facts and retractions get in the way of a good story?
The real story is far less tidy than the LinkedIn version, but far more useful to publishers precisely because it’s less tidy: Iron Maiden built a South American audience the old way – decades of touring, fan-club relationships, promoter ties – and piracy data later happened to describe that audience rather than create the strategy for finding it.
There’s an entire series of essays on cause and effect and misinterpretation begging to be written about the publishing industry. You know the stuff. My book isn’t selling. My freelance gigs are down. I lost my job putting 50% off prices on remaindered books. AI exists, therefore AI made all those things happen. Were would we be without AI to take the blame for our problems?
Which brings us to disclosure number three: I’m a huge AI fan, as regulars here at TNPS know, but that doesn’t mean it gets a free ride. AI is a tool, and it has more than its fair share of faults while being totally devoid of thoughts. If we can grasp that simple concept then AI becomes a powerful collaborative tool.
Back to Iron Maiden.
The more interesting story: what nine AIs did with the same question
In the course of researching this piece, I put the LinkedIn claim to nine freemium AI tools and asked each to verify it, find parallels in other industries, and consider what it might mean for book publishing and for the “AI piracy” debate around training data.
The result was a near-even split on the single most basic, checkable fact in the entire exercise.
Five tools surfaced the retraction unprompted and correctly described the myth as a myth. Three asserted the original, debunked version as settled fact – one stating flatly that the story “is verified and holds up well,” another inventing a specific revenue figure lifted straight from the unretracted viral version.
None of this required deep investigation; the correction has been online, attributed, and uncontradicted for over a decade. My research essentially was to get the actual quotes from the parties involved. But multiple AIs let me down completely. And had I not been a) vaguely acquainted with the 2013 saga, and b) acutely aware AI is not perfect, makes mistakes, and needs treating with caution, this essay might have been an entirely different story about how piracy boosts sales.
But here’s the thing: That AI split matters more than the Iron Maiden anecdote itself. It’s the same failure mode this publication documented when running The Cuckoo’s Calling through six AI-detection tools and getting results ranging from “100% human” to “73% AI” on the identical text.
Ask an AI tool to fact-check a popular anecdote, and you get an answer that depends less on the underlying truth than on which tool you happened to ask. For an industry increasingly leaning on AI tools for research, sourcing, and even editorial judgment, that’s not a curiosity. It’s a liability.
Is piracy theft? What the law actually says
Before going any further with the Iron Maiden story, it’s worth stress-testing the phrase publishing reaches for instinctively whenever this subject comes up. “Piracy is theft” is doing more rhetorical work than legal work, and the gap between the two is bigger than the industry generally lets on.
Theft, as a matter of law, has never required financial harm. The classic definition – codified in England’s Theft Act 1968 – turns on the appropriation of property belonging to another, done dishonestly, with intent to permanently deprive the owner of it. Whether a sale was actually lost has never been part of the test.
Copyright infringement is a different legal category altogether, and the US Supreme Court drew that line explicitly in Dowling v. United States (1985), a case about bootleg Elvis Presley recordings.
Full disclosure four – I have heard Elvis Presley. And fuller-still disclosure – I use video from YouTube of Elvis to show the kids at my school one of the most famous names in music history, and likely as not some of those may be pirate copies. I like to think Elvis wouldn’t mind too much.
Anyway, the Court held that infringing copies are not “stolen property,” on the reasoning that a copyright holder’s property rights are distinct in character from ownership of physical goods, and that an infringer never takes physical control of the work or wholly deprives its owner of the use of it. The slogan “piracy is theft” – coined by an anti-software-piracy lobby group in the 1980s – has been rejected by courts and legislatures more or less since it was minted, for precisely this reason: theft and infringement are not the same wrong, and treating them as interchangeable imports the moral certainty of one onto the economic ambiguity of the other.
A useful real-world test of this: English law’s so-called “Four Fs” rule lets anyone pick wild mushrooms, fruit, flowers or foliage for personal use without it counting as theft – but the exception applies only where the plant is growing wild. It evaporates the moment the fruit is cultivated. Picking apples from a commercial orchard without permission is theft in the strict legal sense whether or not those apples would otherwise have rotted unharvested; the law has never made “would this have sold anyway” part of the definition of the offence.
Food rescued from a supermarket skip sits in similarly unsettled territory – in jurisdictions where discarded stock is deemed to remain the retailer’s property until formally collected, taking it can stay technically unauthorised even though, by definition, it has zero remaining commercial value to anyone. Permission and harm turn out to be two separate axes, not one. “Piracy is theft” only works as a slogan by collapsing them back together.
Jurisdiction complicates things further. Copyright is territorial: there is no single global copyright law, only national laws stitched together by treaties such as Berne, applied under a doctrine known as lex loci protectionis – the law of the country where protection is sought governs. That’s straightforward for domestic disputes; online, it’s contested ground, with courts applying different and sometimes conflicting tests for where an infringing act is deemed to occur – where content was uploaded, where the server sits, where it was accessed, or where it was aimed.
The practical upshot is that a reader outside the country where content was first published is not straightforwardly bound by that country’s law at all – the more relevant legal exposure generally sits with whoever published the material, not whoever read it.
And when it comes to books, reading, in any case, is generally not the infringing act. The UK Supreme Court and the Court of Justice of the EU settled this directly in the long-running Meltwater litigation, ruling that the temporary copies a browser creates while displaying a web page fall under an explicit exception, precisely because the alternative – every reader of every page separately liable for copyright infringement – would make ordinary browsing untenable.
Liability sits with whoever uploaded the content without permission, not with the person who merely viewed it. That distinction rarely surfaces in publishing’s “piracy is theft” commentary, but it is settled law in the jurisdictions that matter most to digital publishing.
None of this argues piracy away, or makes unauthorised redistribution of copyrighted work acceptable. It simply means the industry’s favourite framing borrows the certainty of property law to answer a question property law was never built to answer.
Back to Iron Maiden. And, yes, Game of Thrones.
Why the Iron Maiden logic struggles
Strip away the myth and the underlying business logic is still worth examining – it just needs a harder look than “piracy reveals demand, therefore monetise the demand.”
The reason Iron Maiden’s real strategy (heavy touring, piracy as confirmation rather than cause) worked is that a concert is a rival good. You cannot pirate a stadium seat. Whatever else BitTorrent gave away for free, it could never substitute for the thing the band was actually selling.
The same logic underwrites the other industry examples that hold up under scrutiny: HBO’s tolerance of Game of Thrones piracy worked alongside a subscription product people still had to pay for. Gaming platform Valve’s “piracy is a service problem” thesis converted pirates by selling convenience, not by selling the games for free.
Repeat for video, music and books. But with books it gets complicated.
A book is not a concert. An ebook is the product. Read the pirated PDF/epub and you have consumed the thing being sold, in full, with nothing scarce left to monetise afterwards. That’s the structural reason the academic literature on book piracy is more mixed than the music-industry anecdotes suggest – some studies find meaningful sales displacement for digital editions, others find a discovery effect for backlist or lesser-known titles, and the honest summary is that the effect depends heavily on title, format and market, not a single tidy number.
One underexplored angle: Game of Thrones is always told as a television piracy story, but the underlying IP is a series of novels.
If piracy-as-demand-signal really transfers across media, the test case is sitting in plain sight – did the show’s status as the most-pirated programme on earth move print and ebook sales of Martin’s novels, independent of the show’s own marketing push? That’s a narrower, possibly checkable question, and as far as I can tell, nobody has published that specific cross-media analysis.
Folding in “AI piracy” – and where the analogy breaks down
The instinct to fold AI training-data disputes into this same frame is understandable, but it warrants a more precise legal grounding than “AI companies pirated books, therefore maybe there’s an upside like there was for Iron Maiden,” which seemed to be the narrative thrust this past week.
The relevant case is Bartz v. Anthropic. In June 2025, Judge William Alsup drew a sharp line: training an AI model on legally acquired books was ruled “quintessentially transformative” fair use – that question is settled, at least for now, and not really in dispute in the judgment.
What was not protected was the acquisition method.
Anthropic owned up to obtaining (legally one presumes) a huge volume of printed books which were scanned and then destroyed. This act clearly was covered by the fair use ruling. Anthropic paid for the physical books and the indignation about a big company destroying books has to sit alongside common knowledge that the publishing industry has been pulping books for an eternity. Possibly even longer!
But simultaneously, Anthropic had downloaded close to half a million books from shadow libraries such as Library Genesis and Pirate Library Mirror. That piracy – the sourcing, not the training – is what produced the $1.5 billion settlement approved through 2025 and 2026, working out to roughly $3,000 per affected title.
That distinction matters enormously for this argument. The court did not find that AI training itself is the “Iron Maiden problem” waiting for a touring-style monetisation fix. It found that legitimately licensed training is fine, and pirated-source training is a half-a-billion-dollar-a-year liability with no equivalent of a stadium tour to offset it.
There is no rival good sitting alongside an AI training run the way there was alongside a Maiden gig – no scarce experience left to sell once the data has been ingested. The licensing deals now emerging between publishers and AI firms aren’t a discovery of unexplored upside in piracy; they’re the conventional, court-endorsed alternative to it.
The takeaway
The Iron Maiden story survives because it’s a satisfying inversion of an industry’s standard assumption. And I love me an inversion! But the inversion in this instance is built on a retraction nobody bothered to check, the underlying business logic depends on a rival-good economics that doesn’t carry over to books or to AI training, and the legal framing propping up the opposite argument – “piracy is theft, full stop” – is one the courts themselves have rejected.
On top of all that, the AI tools publishers increasingly rely on to research stories like this one are themselves a coin flip on whether they’ll catch any of it.
If there’s a real lesson in all of this for publishing – apart from treating AI with due caution and doing your homework – it isn’t “piracy reveals revenue you’re leaving on the table,” and it isn’t “piracy is simple theft” either.
Both are doing the same thing from opposite directions – reaching for a clean, moralised story where the actual legal and economic picture is conditional, jurisdiction-dependent, and format-specific.
What’s actually converting unauthorised use into revenue is licensing markets, not myths and not slogans – and that’s true whether the unauthorised use is a torrented album or a scraped backlist.
This post first appeared in the TNPS LinkedIn Analysis newsletter.
