Patent Law Can't Keep Up With AI's Pace Of Change, Creating Massive Risk And Massive Opportunity

The existing U.S. legal framework for the protection of intellectual property through patents and copyrights is groaning under the strain of AI’s massive acceleration of human creativity and invention. 

The cracks that are appearing represent huge opportunities for entrepreneurs and startups – and significant and growing risks for incumbents.

The leverage that AI can provide a startup or even a single entrepreneur, combined with the huge gray areas in IP law, means that new AI-assisted product and process innovations – and possibly even AI prompts themselves – could now be patentable. 

That could create scenarios, given the “first to file” system for patent priority in the US, in which a vast gold rush for patents on AI-assisted inventions starts.

Where it stops, no one quite knows. When it stops is probably only after courts and the Congress catch up. How long that will take, given the speed of change, the many sides with interests tied up in the issue, and the massive financial stakes, is highly uncertain. 

And the amount of money to be made and lost in the process means companies ignore the question at their significant peril.

This isn’t the first time this has happened -- US patent and IP law always lag behind changes in technology. There was much uncertainty around whether software was patentable in the 1960s and 70s, and there were titanic legal battles over biotech and gene patents, as well as protecting the IP aspects of the Internet and e-commerce. 

And there’s little doubt that federal courts – and potentially the Congress -- will, at some point, need to wade into many of these issues to provide definition and clarity. But as in those earlier cases of fast-moving technology, the speed of reality is far exceeding the speed with which the legal frameworks can be remade. 

That’s an invitation for bold action and pushing boundaries. Given the speed at which AI platforms self-improve, this is almost like a warp in the fabric of innovation opportunity.

Since software’s functionality can be patented if it accomplishes an action or approaches a problem in a unique way, the idea of seeking protection for an algorithm whose code was generated by an AI platform in response to a prompt written in English by a human is not far-fetched. 

By offering high-powered, low-cost support in the creation of new IP, large language models can make traditional barriers to entry far less powerful.

If an LLM can replace an office campus full of engineers in the development of software, then a single founder with a particular insight and a knowledge of how to employ an AI platform could conceivably create patentable code that unlocks a new market – or blocks a highly established company from pursuing it in the same way without permission.

So in theory the threat and opportunity level in the intellectual property world is flashing red – or green, depending on which side you’re coming from. 

If a prompt given to a large language model (LLM) describes an innovative algorithm in a general way, and the LLM generates source code that performs this algorithm, then the prompt itself could help secure a patent for the algorithm. That’s because the LLM can quickly create working versions of the algorithm, proving its functionality, which is necessary for patentability -- and do so very quickly. 

Additionally, the LLM enables people without coding skills to demonstrate that their innovative algorithms work, making it easier for a wider range of people to both create working versions of software and to obtain patents on that software.

The speed and ability to innovate becomes even more intimidating as we start to see the early work of “agent-based AI,” where two or more AI systems communicate with each other. One platform gets the prompt and the second can evaluate the prompt or access data to assist the first. 

The platforms can share and iterate – kind of exactly how software engineers working at a company would over weeks or months of pizza-fueled all-nighters. The difference, of course, is the agent-based AI systems can iterate quite literally at the speed of light.

If that’s not scary enough, in this agent-based scenario, the innovator can potentially start out with a high level description of an algorithm – quite short on specifics – and use this type of agent-based AI to create a range of ways of making the algorithm work. 

That could result, under our current patent law framework, in a very broad patent that is also highly defensible. And it could severely restrict the economic opportunity for others in the same field. 

Is that scenario certain? Definitely not. The best case for the prompts being patentable in a case described above is if they seek the creation of code or products that are new and not obvious. 

Another way to understand the impact of AI in this context is that, in the past, when a new invention was created (like the steam engine), it took a long time for the inventor of that technology -- and other inventors -- to improve on that invention and to come up with ways to use it in different types of machines and processes (e.g., steam train, steam ship, etc.). 

That meant that many people typically obtained patent protection over a range of related technologies over some number of years. Innovating was anything but frictionless, and the time, effort, talent and money required for innovation acted as a  safeguard against concentration of innovation and patents. 

But with AI, when someone creates an initial invention, AI makes it increasingly possible for that initial inventor to flesh out different variations of that invention and applications thereof more easily and rapidly. That in turn makes it easier for the initial inventor to get protection for a broad range of inventions surrounding the initial invention more quickly and inexpensively -- and before others have time to obtain competing IP.

This is how AI is both compressing the timescale of innovation and putting immense pressure on everyone to innovate rapidly and to obtain IP protection for their inventions that is as broad as possible, as quickly as possible. It is, in effect, pulling the friction out of innovation.

Open AI founder Sam Altman recently said that he thinks AI can help create a billion dollar unicorn startup with only the founder on its payroll.  Whether that becomes literally true or not doesn’t matter – the idea he’s expressing is directionally correct. A founder who’s savvy in the use of AI could create an astonishingly valuable business with shockingly little capital and very few employees -- and very quickly. 

And under current patent law, erect a legal protective wall around how they did it so others couldn’t follow without paying a toll.

That means no company, no matter how impenetrable its barriers to entry and its IP protections seem today, is invulnerable. 

AI platforms can reduce the time and expense of innovation and therefore reduce the cost of failure. A reduced cost of failure means a vastly reduced cost and speed of iteration. The data from failures becomes grist for the AI mill to help it figure out how to succeed.

This all means that competitors – and invisible one-person startups you didn’t know were competitors – can be studying what your company is up to, determining what your likely next steps are, and using AI as their R&D engine, take away your future. And if those competitors have IP protection for their AI-assisted innovations, you’ll be legally blocked from competing with them. 

Those are the threats. But the opportunity is similarly massive. A company should have better existing data about its market, its customers, its processes, its opportunities, and its technology than a new market entrant would. It also will have existing relationships with customers and a much better ability to sell new and innovative services to them than a new-entrant competitor would. The large company might even have long-term / exclusive contracts with its customers, which can in themselves be a significant barrier to entry to a newcomer. 

Finally, a successful existing company should have greater scale and resources, and be able to put all those together through the magnifying and amplifying lens of AI, to get there first with something better.

AI is raising the bar for what’s patentable, as it in effect makes the way to do new things more obvious, and therefore potentially not protectable. So an existing market participant needs to leverage AI to maximize its inventive skills and help power it past that obviousness threshold. 

The bar, in other words, is going up – but so is your leaping ability. 

The practical and more specific implications for companies are that they need to accelerate the pace of their patent filings. Of course a patent attorney like me would say so – but it’s reality. We’ve got a first to file system in the U.S. and I fear smart innovators armed with AI will be jamming up the filing windows.

We’re entering a time of uncertain boundaries, changing legal frameworks, decreased friction and increased competition. 

The acceleration in the business world from AI has only barely begun – but watch as these jets kick in. The economic G-forces will be remarkable. Don’t sit on the sidelines while others fly the plane.