Intangible assets represent over 90% of the value of every AI company. 

An effective IP strategy can help protect these assets and steer your Company to higher valuations and provide leverage in commercialization of your technology. 

To build an IP strategy that accomplishes these goals, it is important that the focus of IP efforts be on the revenue drivers of your business, present and future. 

While from an IP perspective, an AI business is a lot like any other software business, there are some rather unique differences that require nuance in ensuring that your IP strategy is properly designed. Otherwise, your Company may end up for example with an unproductive focus on patents only or gaps in your approach to IP governance that can be costly or cause risk to your Company. 

Here are a few examples of how these nuances can play out in development of IP strategies for AI technology or platforms. 

1. Data intensity. AI companies are extremely data-driven – as everyone knows. This is especially true when one considers what are or will be the most valuable intangible assets of the Company. Legal protection of data assets is notoriously complex and jurisdiction dependent. For example, in Europe, AI tools that incorporate third party information need to keep track of where the information originated from. Also, AI companies tend to view legal issues related to data through the lens of privacy law, and certainly privacy compliance is critical. However, ensuring that AI companies have the legal rights to exploit their products – including data assets – requires that their approach to data governance consider such legal issues as IP protection available for certain types of data and establishment of legal chain-of-title that supports the legal rights to data that the Company requires to execute on its business plan. This necessitates establishing a link between the Company’s approach to data governance and its overall IP strategy that is often absent.
2. Contracts rule. Possibly more than in the case of any other technology domain, expert drafting and negotiation of important contracts – both those related to the development of intangible assets and those to their commercialization – is vital to ensuring that the Company has the legal right to exploit its IP. In many ways, having your contracts say the right thing about such important issues as ownership of derivative data or technology improvements and the right to use third party software or data to develop and commercialize AI products, is the most important part of a well-thought-out IP strategy for an AI business. And this needs to happen in a way that accounts for business realities including that AI is often developed collaboratively with others who have their own business interests or in a way that relies on 3rd party platforms or data. How the AI company approaches its agreements or terms of use with these counterparties will have lasting impact on the value of the Company. And any gaps often mean an expensive fix. It is also important to ensure that the IP strategy accounts for the varying legal rules governing ownership of contributions to AI tools by employees or contractors, depending on their jurisdiction. For example, contributions to AI technology by employees or contractors in Europe may create ownership rights that are difficult to design around, An IP strategy should map out a plan for navigating the negotiation of these contracts, or how to work around less-than-optimal terms of existing agreements. 
3. Tough-to-protect innovations. In many AI companies, the nature of the valuable intangible assets that provide competitive advantage can be hard to protect using intellectual property filings such as patents. For reasons explained in this section, making the right call on what IP tools to use covering which aspects of the Company’s AI innovations can be challenging, but given an increasingly crowded market for AI innovations and cluttered patent landscape, it is important to dig into the specific circumstances of the Company and develop the right patent strategy. Part of the challenge is that software, algorithms, models, UI/UX designs, and applied know-how together represent substantial value in most AI companies, yet there can be significant roadblocks to patenting such innovations. Add to this that many AI companies use third party tools or known techniques, and their innovation may reside in exactly what techniques are used in combination and how. AI companies can sometimes find themselves in a Catch-22 whereby they disclose substantial details regarding a differentiated implementation in a patent application, which turns out not to be patentable or only extremely narrowly so. Or the claims have limited enforceability[1]. Either way, competitors may end up, in a worst-case scenario, with a quasi-blueprint to practice the Company’s technology without enforceable exclusive rights. Fear of this scenario has likely contributed to some AI companies (and indeed tech companies more generally) investing less in patenting or avoiding patents all together.

This is very often the wrong strategy because there may nonetheless be interesting opportunities to carve out exclusivity by focusing patenting efforts for example to systems that are in effect an implementation use case of the Company’s AI technology or by registering design patents covering innovative design elements of a dashboard that helps drive adoption of the platform. There can be valuable IP land to grab, and if you don’t lay your claim, odds are your competitor will. Being boxed in by a competitor’s patents is a real risk that AI companies eventually face, and mitigating it is generally expensive, time consuming and often results in less-than-optimal outcomes. In addition, you are put in less-than-ideal negotiation position if/when you receive an “offer” letter from a competitor to license their patents that they have filed on similar subject matter. Better then to construct a well-thought-out IP strategy that for example identifies relevant patents for freedom-to-operate and specific aspects of AI technologies as patent targets and associated patent prosecution tactics for patenting that together will deliver substantial value to the Company and mitigate future risks. Such patent filing strategies will generally focus on patent quality over patent quantity, and drawing the line between AI innovations to be patented versus those that will be protected using trade secrets is one of the most important domains where the Company will benefit from expert, business pragmatic IP advice.

1. Trade secrets strategy and management. For reasons covered in points #1 to #3, trade secrets protection will be an important part of any well-managed AI company’s IP strategy. But for this protection to be effective it must intentional and operationalized through carefully designed business processes. For too many AI companies, trade secrets are used as a default IP category with no real strategy or business processes backing it up. This represents a big “gap” revealed in due diligence for example in connection with a financing where the stakes are high enough, or where the Company is in a situation where it needs to enforce its trade secrets. Classically this need arises when an employee leaves to join a competitor or to form one. Conversely, building out processes to identify and manage valuable trade secrets, not just the Company “crown jewels” but also important categories such as applied know-how – with proper guidance – can be efficient and effective. Trade secrets management can include development of a trade secrets register to categorize trade secrets and log appropriate protections that mitigate risk but don’t bring development and commercialization to a grinding halt. Most clients who take on design of a proper trade secrets management program report that it is not a “heavy lift” and that it delivered improved risk management and helped them justify higher valuations with stakeholders. 
2. Open source and open data. Use of open source and open data is particularly common in AI. However not everything that is “open” is truly as open from a legal point of view as it may seem. Use of some open-source software for example in some ways, can be “fine” – or not – depending on the fine print of the terms. The Catch-22 in open-source software is use of software subject to problematic terms (e.g. requiring disclosure of source code, limiting commercial use, and/or restricting IP protection of broadly derivative works) in connection with an innovation that is core to the Company’s competitive differentiation. Remediation is usually possible but can be very costly and takes time. If these issues are identified in the middle of legal due diligence, it can be a pretty big problem in the deal. Conversely, some basic controls around what types of open source or open data are used (depending on the risks associated with their terms) and how, can avoid many of these problems. All too often it turns out that there was a close alternative to using an open-source library with problematic terms, one with relatively friendly terms.

In conclusion, all this and other nuance is resolvable by AI companies, with disciplined commitment to work towards a value-driven IP strategy that also mitigates risks. With expert guidance, one can focus efforts and budget on building IP assets that will matter and scale up these efforts at the appropriate point in the development of the business. 

[1] A key consideration in developing a patent strategy for AI technology is ensuring that patent claims are enforceable by applying up-to-date guidance based on the evolving interpretation by U.S. courts of the tests for determining patentability of software patents originally set out in Alice v. CLS Bank.