Artificial Intelligence and Intellectual Property – Considerations for Businesses
The use of artificial intelligence (“AI”) has sky-rocketed in recent times. Most people are aware of publicly available AI language models such as ChatGPT, Bing Chat, or Google Gemini. However, there are also numerous commercially available AI solutions for use within businesses. Companies are increasingly keen to embrace the potential time- and cost-saving benefits that come with using AI, however, companies should also be mindful of potential issues that may be caused by it.
The relationship between AI and intellectual property (“IP”) is complicated, and legal uncertainty has been created by the upsurge in AI usage over the past decade. It is a feature of the modern world that technology progresses extremely quickly – often outpacing legislation – and the mismatch between AI tools and IP law is an example of this. When considering whether to implement an AI model within its business operations, a company should be mindful not only of general commercial risks, but also of legal risks relating to aspects of IP law, such as copyright and trade mark. Some of these are discussed below.
Copyright and Trade Marks – A brief overview
Copyright
In general terms, copyright gives an author, artist, creator, or composer, of certain works the legal right to prevent others from copying or exploiting their work. This right exists as soon as a work is created and written down or recorded (whether on film, by sound recording, or on the internet). Broadly, where a third party copies, issues copies of, or communicates copyrighted work to the public, without the owner’s permission, they infringe the copyright owner’s IP rights.
Trade marks
A trade mark is any ‘sign’ (this could include words, slogans, designs, or letters) used by a ‘trader’ to differentiate its goods from those of others. A trade mark may communicate, among other things, a brand’s story or the quality or provenance of its products, and therefore a trade mark can itself be a highly valuable commercial asset. Broadly, IP rights will be infringed where a person or company uses a sign in the course of trade, without consent, that is identical or likely to be mistaken for the trade mark of another party.
Types of AI Platform
AI platforms can be divided into two categories: ‘cloud-based’ or ‘on-premise’. ‘Cloud-based’ platforms are those for which data is hosted externally on network cloud storage, rather than within the physical confines of the user company, whereas ‘on-premise’ platforms are those providing AI services that are hosted ‘in-house’, i.e. contained within a company’s physical environment rather than being reliant on external infrastructure.
When assessing which type of AI platform to implement, a company should consider the platform’s technical design in the context of IP law. Although cloud-based platforms may provide flexibility and cost-efficiency, they increase the risk that data entered will be uploaded or shared externally by the AI provider. On-premise platforms, on the other hand, give a company control over the data entered into and hosted on the platform, therefore, whether data is shared with third parties is a decision for the business.
In light of this, a company should carefully consider the proposed contractual terms between it and the AI provider. If those terms are silent as to whether the data input can be stored or shared externally, the company should ensure its users avoid entering content that could be the subject of IP rights (whether its own or that of its business associates), including copyrighted works, trade marks, or trade secrets. Otherwise, there is a risk that internal IP-protected works will inadvertently fall into the hands of third parties.
AI training – Input and Output
For AI models to work effectively, they must first be ‘trained’. AI training typically involves feeding a considerable quantity of data into the AI model. Where large amounts of data are taken from publicly available websites, this is known as ‘web scraping’. The AI model then uses the training data to generate content and perform tasks as prompted by the user.
Users are not typically aware of the nature of data used to train an AI model, nor are they likely to know whether the owner of the IP rights to any material used has consented to its use. This creates the risk that the data used to ‘train’ the AI model included content protected by IP rights. Therefore, there exists the possibility that the AI model generates content that contains elements of IP-protected works. If the AI output is shared externally, this could leave the company vulnerable to legal challenges for IP infringement. Further, it is unlikely that an AI provider will offer to accept liability for the IP infringements of those using its platform – in fact, its contractual terms may expressly exclude this – therefore, the obligation to mitigate this risk falls on the users.
Considerations for businesses
The popularity of AI is growing rapidly and its widespread use within businesses in the near future seems inevitable; commercial users should be mindful of the risks associated with embracing AI. The relationship between AI and IP law is complex and evolving but, as a starting point, companies should carefully consider the following:
1) Type of platform – Is the AI platform ‘cloud-based’ or ‘on-premise’?
2) The commercial contract – Does the contractual agreement between the company and the AI provider allow for external storage or use of inputted data? Will the AI provider accept any liability for IP infringements (or other legal issues)?
3) AI training – The AI platform may have been trained using data obtained through ‘web scraping’, and the possibility exists that the data used may have contained IP-protected works.
4) AI output – Does the content produced by the AI platform contain anything that may infringe on third-party IP rights? If so, external use of the content should be approached with caution.
5) Policies and staff training – In the absence of reliable automated methods to check whether AI output contains IP-protected works, it may be sensible to implement an AI usage policy and provide basic training to staff on the advantages and disadvantages of commercial AI use.
Should you have any queries in relation to the commercial use of AI, please do not hesitate to contact our Commercial Team who would be happy to assist.