AI and Intellectual Property

Do You Really Own Your App?

Depending on who you listen to, Artificial Intelligence (AI) is either a boon for mankind or, according to Elon Musk, the greatest existential threat we face as a species. It’s the “dangers” of AI that have fulled most regulation currently being discussed, with news reports focusing mainly on the perceived risks for citizens’ fundamental rights and freedoms. However, there are also legal liability concerns for both developers and users of AI inspired applications and tools.

AI and Intellectual Property Issues

In particular, the development and increasing use of generative AI systems – the technology that enables the creation of content on demand – does have significant potential, but also raises significant legal issues. Upcoming regulation has relatively little to say so far about the complex – but highly commercially relevant for many tech start-ups – intellectual property (IP) implications of AI systems. A situation made even more complex by the cross-jurisdictional nature of online data scraping and use, and territorially different intellectual property laws.

AI Ownership and Copyright Issues

A start-up tech firm or an individual who has developed an app using AI may choose to monetise it by placing it on a major player’s platform and it’s important that the developer can say they own the IP in the app. Establishing this if AI has been involved in the development is not as clear cut as you may imagine.

The UK’s position on copyright ownership for AI generated works remains ambiguous, while other jurisdictions such as the USA, Spain, and Germany have clarified theirs, stating that copyright can only exist and be owned by a human being.

Is Your AI System Protected?

Some aspects of Al systems can be protected by patents and copyright, but machine learning systems don’t fit neatly into conventional IP categories. This is in part due to the fact that the real investment and value in developing a generative Al solution goes into the training process, which makes the system inherently much more valuable. However, generative AI relies on the input of large amounts of training data, which may or may not be subject to IP restrictions. Photos and text underlying some of the popular generative AI applications are likely to be copyright protected, but whether the training set as a whole is protected varies depending on the jurisdiction (the European Union for example has a specific database protection right).

Will The Output Infringe Copyright?

The output of a generative AI system will inevitably contain traces of the training input, and these can sometimes be readily identifiable. For example, visual media giant Getty Images filed a case earlier this year in the High Court of Justice in London against Stability AI, alleging that the company copied 12 million images to train its AI model without permission or compensation.

Stability AI and another AI art start-up – Midjourney are also being sued in the US by artists Sarah Andersen, Kelly McKernan, and Karla Ortiz who claim that these organisations infringed their rights by training their AI tools on images scraped from the web without the consent of the original artists.

In the US, “fair use” can be a defence against claims of copyright infringement. The digitisation of copyrighted books for an online library project has been interpreted as “fair use” by US courts but it is not yet clear whether this exception will apply in the context of AI. Most European copyright laws do not recognise “fair use”, so it’s more difficult to justify the use of a third party’s copyrighted works in the input or output of an AI system. If third-party copyright is infringed, it’s the user and the AI developer who would be liable.

Avoiding Costly litigation

To avoid the possibility of costly and potentially business-breaking legal problems, AI developers should ensure they comply with the law when acquiring training data. This could be done through licensing agreements or by sharing revenue generated by the AI tool with the individuals who own the IP of that training data.

Equally, customers of AI tools and apps are increasingly likely to ask providers and developers whether their models were trained with any copyright protected content. To avoid risk to themselves, they will shun generative AI tools that cannot confirm training data was properly licensed from content creators or subject to open-source licenses with which the AI company complied.

Our team of start-up lawyers at Motion Paradox can give AI developers and corporate users of this emerging technology guidance in navigating a largely unregulated landscape. We can ensure your business is protected from legal issues that may damage your company.