A new report published in August by the Center for Data Innovation found that while the U.S. was still leading the worldwide race in AI innovation, China could be close behind. This is in large part, it found, because of easier access to the data sets needed to train AI models and an influx of venture capital and private equity funding that already places it ahead of the European Union.

But U.S. companies also have to contend with the stringent legal framework around patents for AI software, which in some cases can make it difficult for companies or inventors to obtain protection against infringement.

"Fundamentally, I believe one driver of our entire economy is the limited monopoly that inventors get through the patent system. So if you remove that protection for AI, that adds another layer of potential advantage for a Chinese company or anyone else in the world for that matter to come in and compete," said Thomas Isaacson, a shareholder at Polsinelli.

One of the complicating factors that companies face in attempting to patent a piece of AI software is a 2014 Supreme Court decision in the case of Alice Corporation PTY. LTD. vs. CLS Bank International. CLS Bank had filed a suit against Alice Corporation alleging that patents the company held in relation to a computer system that behaved as a third party intermediary between two parties exchanging financial obligations were "invalid, unenforceable and not infringed." Alice Corporation alleged infringement in a counterclaim.

However, the Supreme Court ruled that the claims were based on a "patent ineligible abstract idea."  Per Isaacson, software can run afoul of the framework established here if it sounds too much like a basic computer performing a series of steps that a user could perform in their head—which is sort of the point behind AI.

"If you think about it, artificial intelligence is basically getting a computer to think like a human.  That is a hugely complicated process but can come across as not patent eligible under a loose application of Alice," Isaacson said.

Therefore, drafting a successful patent application will most likely entail robust care. While some corporate legal departments may opt to approach the process themselves, there may be value in seeking additional support.

Unfortunately finding that special someone isn't always easy.

"Sometimes clients are having difficulty finding people with expertise at law firms. From a law firm perspective we still think there's a lot of opportunities for protecting AI innovation through patents," said Carl Kukkonen, a partner with Jones Day.

Successfully executing those opportunities may come down to the way a patent application is framed. For starters, highlighting a specific problem that a piece of AI software solves or identifying a manifestation in the physical world (Isaacson used the example of controlling a mechanical arm) can help a company's chances.

Kukkonen also thinks it will become important to tie applications into technological improvement like making a computer process work faster or be more efficient storage. Software geared towards user experience is harder to get approved, but that doesn't mean that companies have given up on securing AI patents altogether.

"We still see that companies are putting a lot of time and effort and investment with regard to AI-oriented patents," Kukkonen said.

But what happens when those efforts fail? Sending a product into the world without a patent is risky and may also force a company to place a greater emphasis on the protection of trade secrets.

Kukkonen suggested that legal departments could assist in deploying in-house policies for identifying trade secrets.

"If they ever do have a leakage of some of their trade secrets, the fact that they've identified and catalogued their trade secrets in advance of such a leakage will make enforcement potentially easier down the road," he said.