Authors
A recent Third Circuit decision just handed a significant win to anyone who believes the law should be freely accessible. The ripple effects will be felt beyond building codes, potentially influencing how courts think about AI training on copyrighted content.
Copyright versus public access
Private organizations write technical standards: detailed rules about everything from how steel beams should be manufactured to how electrical systems should be wired. These standards are copyrighted, and selling them is big business. ASTM, the plaintiff here, generates about 70% of its revenue from standards sales.
The complication
Governments will often adopt these private standards as binding law. When a city building code says “steel pipe piles must comply with ASTM A252,” that standard isn’t just a best-practice recommendation anymore, it is the law. At that point, it seems odd that you might have to pay ASTM for access in order to read that law.
Enter UpCodes, a company that built a searchable online platform to help people access building codes and the standards incorporated within them, for free. Their pitch: if it’s the law, you shouldn’t have to pay to read it. They post only versions of standards that have actually been adopted into law (usually not the latest editions), organize them by jurisdiction, and include any local amendments.
ASTM disagreed, strongly. They sued for copyright infringement and asked the court for a preliminary injunction to force UpCodes to take down the standards. The district court said no, finding UpCodes was likely to win on fair use grounds. ASTM appealed, and the Third Circuit just affirmed.
How Fair Use Works (The Short Version)
Think of fair use as a safety valve for copyrights. The doctrine recognizes that sometimes copying, even without permission, may serve the public interest. Courts weigh four factors: (1) the purpose and character of the use (is it transformative?), (2) the nature of the original work (factual or creative?), (3) how much was copied, and (4) the effect on the market for the original. No single factor is decisive; courts look at the whole picture.
Here’s how the Third Circuit applied each factor:
Factor 1 – transformative purpose: The key question is if the new use serves a different purpose than the original? Here, ASTM creates standards to provide technical guidance for industry professionals, telling engineers and manufacturers how to build things safely. UpCodes uses those same documents for a completely different reason: telling citizens and regulated entities what the law requires. Same text, different mission.
Transformation is about purpose, not alteration — you don’t have to add commentary or change the words to be transformative, just use them for something fundamentally different. The practical point: you can’t provide access to the law without reproducing it.
UpCodes is a for-profit company that sells premium subscriptions. But it gives away access to the standards themselves for free. The premium features are things like bookmarking and AI tools. The court found this “freemium” model meant commerciality didn’t weigh heavily against fair use, since UpCodes wasn’t directly monetizing the copied content and found Factor 1 in favor of UpCodes.
Factor 2 – Nature of the Work: Copyright protects creative expression more strongly than factual or functional content. Technical standards, which specify things like acceptable wall thickness tolerances, are very much on the factual/functional end of the spectrum, and favor fair use. The court added that once standards become law, they move even further from copyright’s core since laws need to be accessible to everyone. Another one for UpCodes.
Factor 3 – Amount Copied: UpCodes copied everything: the full standards, including “non-mandatory” appendices, notes, and tables. Usually, copying a whole work weighs against fair use, but the court found full copying was reasonable here. Why? Because building codes often incorporate entire standards by reference, and you can’t understand or comply with the law by reading only part of it.
Take one example: the ASTM A252 standard on steel pipe piles. The code requires compliance with “the material requirements in ASTM A252” without specifying sections. The standard’s main text sets general rules, but a “non-mandatory” appendix contains the table you actually need to calculate minimum wall thickness. That supposedly optional material is essential for compliance. Context matters. 3-0: UpCodes.
Factor 4 – Market Harm: This is the tough one — ASTM argued that free access to its standards obviously hurts sales. And that makes sense. If you can get something for free, why pay? The court acknowledged some substitution was likely for these specific standards, but ASTM couldn’t show how many subscriptions were actually lost, or what portion of its revenue depends on selling incorporated (often outdated) versions versus current editions. The record suggested that once standards are updated, demand for older incorporated versions drops sharply, which also makes sense. People in the industry want the latest guidance, not last decade’s code requirements.
Weighing public benefits, on the other hand: citizens shouldn’t have to pay to read laws that bind them. Easier access helps contractors comply, helps inspectors enforce, and helps journalists and advocates scrutinize and debate public policy.
Final score on Factor 4: about even, slightly leaning UpCodes.
Bottom Line
The Third Circuit’s ruling is a win for legal transparency. When private content becomes public law, copyright’s grip loosens. This decision offers some useful language for AI stakeholders on transformative purpose and full-text copying, but the analogies only go so far. Legal access to incorporated standards is a narrow use case compared to training commercial AI on creative works. This decision is not a green light for AI training. Fair use remains fact-intensive and context-dependent. This ruling turned on a particular record, a particular purpose, and particular gaps in the evidence and could change through trial as the record develops.
Authors