In the past, the effort invested in building something could often be paid back as a right to stop others from copying it. Development effort still matters. But it is not the strong barrier to competition it once was. A software feature that took a year to develop and refine can now be approximated in weeks by a competitor who never sees your source code, using your own public descriptions of the feature and a generative AI model. AI models are very good at generating code from a detailed description. Whether that competitor faces legal hurdles does not depend on how hard you worked; it depends on which kinds of intellectual property protection you can bring to bear. Three alternate versions of the same story below illustrate why copyright and trade secret by themselves can fall short, and how to improve your options to protect the results of your hard work.
In each version, a company I’ll call Innovator Inc. develops the same software feature and faces the same competitor. Only Innovator’s protection strategy changes, and the outcome changes with it. These are hypotheticals to illustrate differences between intellectual property strategies; your results may vary.
Not Good: Relying Entirely on Copyright and Trade Secret
Innovator Inc. sells software. Over eleven months, its team internally developed, tested, and refined a New Feature, then released a new version of the product that included it. Before release, Innovator’s in-house counsel had suggested the company consider patent protection. Management declined: “Patents are expensive. We already have copyright and trade secret protection. That’s enough.”
Two months after release, a competitor CopyCat Corp. shipped a competing product with a Nearly Same Feature. CopyCat did not break into Innovator’s systems or hire away its engineers. CopyCat’s developers fed Innovator’s marketing materials, user manual, user reviews, and other public descriptions of the New Feature into a generative AI model, asked the model to produce code implementing that functionality, and integrated some of the resulting code into the CopyCat product.
Innovator’s management wanted to sue, but they were advised that the available grounds for suing were very weak. Copyright protects a developer’s particular expression such as the source code Innovator wrote. Copyright does not protect the ideas, processes, or functionality of software. CopyCat never copied Innovator’s code; the AI generated code without accessing Innovator’s code. Trade secret protection covers only what is actually kept secret. Everything CopyCat used was public, so there were no trade secret violations.
The eleven months of effort produced a competitive head start, but only for the two months it took CopyCat to generate the Nearly Same Feature. The functional core of the New Feature was outside the copyright and trade secret protections, basically free for the taking, and AI made the taking relatively fast and cheap. Copyright and trade secret protections still have value, but they were not intended to stop someone from copying functionality, as long as they do it without copying source code and without stealing confidential trade secrets.
Patents are the one kind of intellectual property protection that might have helped Innovator. A patent does not rely on any single expression of functionality, and it does not depend on secrecy (in fact, patents are granted in a bargain by which innovations are made public in exchange for a limited period of exclusivity). A patent protects a claimed invention against anyone who makes, uses, or sells the patented invention, including a competitor who developed the same functionality independently and never saw a line of your code. Copyright has an independent-creation defense; trade secret has a reverse-engineering defense; a patent has neither. Patents that are valid, enforceable, and infringed close that protection gap.
Better: A Patent Application on File Before Release
Replay the story. This time, before releasing the New Feature, Innovator works with patent counsel to identify aspects of the feature that could be protectable by a patent. Not every piece of functionality is patentable. Patent applications face prior art challenges; the invention must be sufficiently different than what has been done before. Inventions involving software or AI also face tough subject-matter-eligibility hurdles; protection often requires describing a specific technical improvement rather than merely a desired result. In this replay, Innovator files a patent application six months before the New Feature is released, and then marks the released product “Patent Pending.”
That notice can change competitor behavior even though no patent has issued. One competitor, call it CautiousCat, sees “Patent Pending,” weighs the uncertain scope and timing of a possible future patent, and decides to take its next release in a different direction instead of trying to reproduce the New Feature.
CopyCat is less cautious. After internal debate, it decides to extract as much benefit as it can and to deal with any infringement allegation if and when one arises. It again uses AI to implement a Nearly Same Feature. In this replay, Nearly Same Feature is released about two months later than in the first version of this story, because the debate over possible patent exposure slowed down the decision-making and the implementation of Nearly Same Feature.
A pending application is not an issued patent, and “Patent Pending” confers no present right to stop anyone. But it can have value. It signals that enforceable claims may be coming, which can deter risk-averse entities such as CautiousCat. It can impose real costs and delays on those who proceed anyway, such as CopyCat. In this version, Innovator is better off than in the first version, because Innovator filed for patent protection before releasing New Feature, and used that application as a legal basis for marking the product as “Patent Pending”.
Best: An Issued Patent, with Options Preserved
Replay one more time. In this version, Innovator files a patent application and also requests accelerated examination of that application. Acceleration does not always produce a patent quickly, and it does not guarantee the desired claims, but in this case it works. Innovator obtains an issued patent eight months after the New Feature was released, with strong claims covering a key part of the New Feature. After seeing the issued patent and consulting its own counsel, CopyCat removes the Nearly Same Feature from its competing product.
An issued patent with enforceable claims changed CopyCat’s own calculation: continuing to ship the Nearly Same Feature was no longer worth the exposure, so CopyCat decided to pull it. Innovator did not have to bring a lawsuit to reach that result. The existence of the patent right was enough to motivate CopyCat’s decision.
To keep its options open, Innovator also files a PCT application, preserving the ability to pursue patent protection in other countries. Innovator also files a continuation application, preserving the ability to pursue additional claims from the same disclosure as competitors and products evolve. Functionality developed as part of those eleven months of Innovator effort is now protected by patent. Innovator also retains flexibility to adapt that protection over time.
The Pattern
The three versions differ mainly in what Innovator did before it released New Feature. The biggest benefits came from filing early for patent protection, after evaluating whether and what to patent, in view of prior art and eligibility, before disclosing the feature to the world.
Generative AI did not change the law in any of these versions. Copyright still protects Innovator’s source code. Trade secret still protects valuable secrets whose disclosure is not required as part of the patent application. What AI changed is the practical reality: copying the unprotected parts of a product is now so fast and inexpensive that relying on a head start, or on an assumption that effort alone deters competitors, is riskier than it used to be. The protections that come with an independent-creation or reverse-engineering escape hatch are easier than ever for a competitor to walk through.
None of this means a patent is always the answer. Patents are not always available, they are not always worth their cost, and other forms of protection sometimes fit better. But they should be considered as possibilities whenever innovative functionality is part of a product or service. Consult counsel who can evaluate what is patentable in your situation, where the eligibility and prior-art risks lie, and how filing, examination, and continuation strategy fit your business goals. Do that before a public release limits your patent options and invites the kind of AI-assisted copying described above.