TPN Europe 2019: Finding the best protection for AI inventions
Companies must consider both patents and trade secrets when looking to protect their artificial intelligence (AI) inventions, according to in-house counsel at Canon and IBM at Technology Patent Network Eu rope, hosted by WIPR, in London this morning, June 20.
Maria Anassutzi, lead European IP counsel at Canon Europe, said that often, the best protection is “an overall IP strategy” that uses a combination of both patents and trade secrets.
Discussing the advantages and disadvantages of each form of protection, Anassutzi said that while a patent may be a more suitable form of protection if the invention is able to be reverse-engineered by competitors, once a patent application is filed, the invention is then disclosed to the general public.
“If the application is rejected, or the patent is granted but later invalidated, companies will potentially lose valuable information that was secret up until that point,” Anassutzi said.
She added: “Another thing companies must consider is the complexity of the technology behind AI inventions, so an inventor must be sure that they will be able to detect when their patented technology is being used by competitors.”
However, when an invention is difficult to reverse engineer, trade secret protection is probably a better choice.
“An advantage of a trade secret is that as long as it remains a secret, the protection is indefinite,” she said. On the other hand, in the EU, patent protection lasts for 20 years.
But on the flip side, Anassutzi added: “Once a trade secret is disclosed and the confidentiality is gone, that’s it. You lose your protection and there’s nothing to go back to.”
One way to obtain the most protection for your IP is to combine both patents and trade secrets.
“When you file for a patent, you can keep certain confidential information out of the patent application. Then you have something which is still a trade secret,” she advised.
Also speaking at the session, Kevin Fournier, IP counsel at IBM in the UK, said that when it comes to AI inventions, companies often file very detailed patents.
“You can easily sail through the patent office when you file such a long, detailed claim because the examiner cannot find prior art,” he said. “But if they can’t find prior art, are you going to be able to find infringement?”
He said that when legal counsel attempt to prove infringement, they often draw up a claim comparison chart.
“You put elements of your patent claims on one side, and the infringing elements of a competitor’s products on the other side,” Fournier explained.
But, he warned, if a company’s patent claims are very long and extremely detailed, and this is further heightened by the complexity of the nature of AI-related inventions, proving infringement can be a difficult task.
“You have to wonder if your patent has any value if you never know if anyone is using it,” Fournier said.
He added that one advantage of protecting AI inventions with trade secrets is that they are often difficult to reverse-engineer because the software used to create the invention can be stored in an internet-based cloud.
“On a cloud, third parties can’t access your software and play with it, so they can’t reverse engineer it,” noted Fournier.
Good practice
The management of trade secrets within a company was also discussed.
Fournier said that it is good practice to make sure that anybody who has access to confidential information or technology is placed under a strict obligation not to disclose or discuss that with others.
Additionally, he said only people who need to see the trade secrets should be given access to them. At IBM, he said the company recently prohibited the use of USB memory sticks.
“The ban on memory sticks was put in place because of how easy it is to lose them. You can easily fumble for your key and the USB can fall out onto the floor,” he said.
Additionally, Anassutzi said measures need to be put in place to differentiate between what is classed as a trade secret, what is classed as confidential information and what is classed as highly confidential information.
Anassutzi said that, at Canon, this is “one of the most important and complicated” things to consider in the management of internal processes.
Fournier also discussed how companies can avoid accusations of infringement when collaborating with third parties on an invention.
For example, if a company works with a third party on the development of an AI invention, its developers may “see something competitive like the source code” of that company’s AI.
When the product developers return to their own company, they may unintentionally use something they picked up from the third party in the development of their own product.
“They could easily be accused of developing a product that is contaminated with the information they saw in the other company’s product,” Fournier said.
“That’s a key thing. You collaborate with somebody else and you want to bring the best of what you both have, but you have to be careful that what you see of the other side’s [information] is not going to hurt you when you go back,” he added.
Fournier said this is why it is important to have a contract in place which addresses these issues.
Another way around this, he said, is to have separate product developers which work on collaborative products and develop solely in-house products.
Did you enjoy reading this story? Sign up to our free daily newsletters and get stories like this sent straight to your inbox.
Today's top stories:
UKIPO invalidates ‘Carry On’ film TMs
