Streamlining relationships with external counsel and disclosing data sets adequately are vital to patent prosecution for artificial intelligence-related products in Europe and the US, according to financial services firms.
Speaking at the Protecting Innovation in the Financial Services Industry conference in New York last week, counsel at Equifax, Morgan Stanley and American Express said AI had become more important to their businesses and, consequently, their overall filing strategies.
“AI is getting pretty exciting because there are lots of use cases taking place in our various businesses,” said Larry Bromberg, global head of IP at Morgan Stanley in New York.
He explained that his firm uses chatbots for factual extraction to improve communication between the firm and clients. Morgan Stanley has also developed AI to tackle cybersecurity challenges by identifying abnormalities and conducting other security-based processes that would otherwise have to be done manually.
“We can use AI to improve those processes and allow people to do their day-to-day jobs better,” he said.
Elizabeth Lester, assistant general counsel IP at Equifax in Atlanta, added that her company uses AI to better analyse risk, including by differentiating between customers with thin and thick credit files to get customers access to credit more easily.
The firm also uses machine learning solutions for data matching to pair the right consumer information set with what it has received from a bank or a contributor.
“It can be challenging to make sure the John Smith you are looking at is the John Smith you should be looking at,” she said.
Equifax launched the first regulatory-compliant machine learning-based credit scoring system last year. The program, called NeuroDecision, was built to improve performance and accuracy and enable businesses to better assess risk.
Maxine Graham, vice president and senior IP counsel at American Express, added that her firm was generally looking to AI to automate processes in ways that would benefit consumers.
External streamlining
But with this greater focus on AI comes the challenge of harvesting invention data from software engineers and meeting the disclosure requirements of the European and US patent offices.
Bromberg at Morgan Stanley said his firm tries to ensure that the right external counsel is put in charge of drafting patents for the right project and with communicating with engineers involved with said project, to achieve proper disclosure.
“Having a partner, associate or firm with experience helps because I want to make the process as streamlined as humanly possible,” he said.
It is annoying for inventors and generally inefficient if they have to explain the basic underlying technology to a patent attorney, he pointed out.
“I have found that disclosures usually do go pretty smoothly, even in the AI space, so long as external counsel are aware of the technology and know where to press the inventors for information at the judicial level.”
Explaining the process that the business goes through before it gets external counsel involved, Bromberg said his team will usually get an email from an inventor informing them of an innovation and will often have already filled in a disclosure form.
Morgan Stanley makes its disclosure forms readily available from its intranet site, which also contains information on general patent and IP and filing practices.
The IP team will then have a phone call with the inventors, who will describe the invention or set up a demonstration to present its features. A prior art search is then conducted to give both teams a chance to talk about what is already out there and work out the differentiating points between what they have seen and what they have invented.
“Once we get it settled that we have one or more points of significant differentiation, the matter is elevated to tech management, who will assess the importance of the invention,” said Bromberg. “If we get the thumbs up, we will engage outside counsel.”
He added that this process becomes easier over time as more inventors go through the process and become familiar with it. Businesses may also find that they develop ‘pockets of innovation’ when they get serial inventors, and those parts of the company quickly acclimatise to the nuances of the patenting process.
Graham at American Express and Lester at Equifax said that giving inventors the opportunity to talk to outside counsel is invaluable to the disclosure process because the team often uncovers more detail about the invention that the inventors do not think to disclose themselves.
“That process allows us to pull out more information and find more points of novelty,” said Graham.
Lester added that her data scientists that work on AI are often more IP savvy than her other employees covering patents. Their knowledge of the patent system varies depending on their backgrounds, she said, but many of them like to write white papers for customers that the business can use as a starting point for applications.
These points, speakers said, were also important to consider when thinking about European applications. Lester said she keeps an eye on international developments and makes sure that her external counsel, which she relies on heavily for drafting, take European filing practices and strategies into account.
“The approach in Europe will be slightly different because you are focusing on computer-implemented inventions,” she said.
Data dilemmas
One of the main challenges with AI disclosure that emerged during the conference was how to manage data. Bromberg pointed out that a lot of machine learning models are trained on data sets that are obtained from open source forums. To mitigate the risks associated with open source, the business has developed a process for when someone wants to bring in these sets.
But “bumps in the road” can emerge during this process, he added, because of the difficulties in identifying associated licences.
“I may try to figure out the licences that are attached to these data sets only to realise that they are covered by an Apache licence and that there is no licence for the data; which means the contributor probably got it from some website and had no idea what rights were associated with that – and that introduced some risk,” he said.
Bromberg also explained that data disclosure presents a unique challenge to AI patent applications. Patents are intended, of course, to give someone enough information about an invention to allow them to replicate it without undue experimentation, but machine learning inventions are not usually replicable without the right data.
That fact raises the question of what the business should disclose; whether attributes of the data by themselves are enough or whether the entire data set needs to be included.
Bromberg said: “This is one of the key areas that counsel know to discuss with data scientists; how do we describe the reading pattern and pick the right data set? A lot of thought goes into drafting an AI patent application for this reason.
“How do we get the right bits in there without disclosing all of your secret sauce?”
The question of whether the business will need to divulge too much information to get a patent then raises the question of whether patent protection is right for a particular AI-related invention or whether it should consider keeping the invention as a trade secret.
“You can face a lot of uphill battles on this question,” said Graham at American Express. “But if you do not have the option to keep the invention as a secret for whatever reason – which is really the only alternative – and it is of value to the company, you have to take a shot at getting a patent.”
She added that inventors are often most resistant to giving away too much information because they do not always understand how much they will have to disclose.
“They will be the ones to pull back and say: ‘No, that is too much.’”
Technology in the financial services industry is developing at an unprecedented rate, and AI is playing a huge part in that development. There are inevitable challenges to protecting these machine learning inventions, but as the technology becomes increasingly important to financial services firms, they will get better and better at it – with a little help.
