Everyone from Fortune 500 companies to small business owners around the world are rapidly embracing the power of AI to increase efficiency, productivity, and innovation. 

But as the business world rushes to adopt AI technology, there are substantial challenges regarding the legal risks of artificial intelligence that must be identified, understood and addressed. 

In today's blog, we'll look at some of the important risks involving AI: privacy, data security, and intellectual property. 

Understanding the Legal Risks of Artificial Intelligence

The advancement and adoption of AI technologies in business has moved far faster than the legal system possibly can. While Europe is in the process of implementing the AI Act, there are no uniform laws governing AI privacy or data security in the United States.  

By their very nature, AI systems require massive amounts of data to train the AI model. This data comes from a dizzying array of sources and often includes the personal identifiable information of customers, such as names, address and social security numbers. 

Even medical records, genetic information, and financial data are used to train AI. The sensitive nature of much of the data that AI is trained on makes AI data security a significant concern.

Not only that, but some types of AI, specifically large language models (LLMs), are trained on vast amounts of textual data including books, articles and web pages throughout the internet. 

Some of this information is subject to copyright, trademark, and patent protection. The question arises whether the use of this data to train the AI model is a violation of those intellectual property rights. 

And because of the way these forms of AI learn from the data they are trained on, query results sometimes closely resemble copyrighted works, raising separate AI intellectual property infringement risks.

These are just a few examples that illustrate the complex legal risk associated with using artificial intelligence for businesses. 

Let's look at each of these areas more in depth.

Artificial Intelligence Risks To Privacy Rights

Because of the vast amounts of data needed to train AI, there is a necessary concern about how personal data is handled. 

Organizations that collect, store, and use personal data to train their AI have a responsibility to ensure that data is kept private - but AI algorithms can be used to search for patterns within the data that can potentially be used to identify individuals and make inferences about their habits, preferences - even their beliefs and health conditions. 

Even when anonymized, this data can still be used to identify individuals by combining multiple sources like GPS location, shopping history and internet activity - all without the individual's consent. 

What's more, there is also great concern that AI's may reflect societal bias. For example, AI systems used in the hiring process can unfairly favor groups from specific backgrounds. Similar issues can arise when AI is used for loan applications and other sensitive areas.

With respect to these issues, explainable AI (XAI) serves an important role. XAI refers to the process and techniques humans use to understand how AI arrives at its decision. XAI is critical to detect biases in AI, as well as debug and improve the technologies. 

Understanding how AI reaches a decision also helps improve trust in its decisions, and ensures that those decisions do not engage in discrimination or biased results. 

The Legal Risks of AI and Data Security

Big data powers AI, but as discussed above that data often contains sensitive personal and health information. That makes security a major concern. 

Some of the major AI and data security risks include:

Data Breaches and Hacks

Threat: AI systems are vulnerable to traditional security breaches and data hacks like any other system. 

Danger: An AI data breach has the potential to expose private user data, trade secrets, and also cause outages of the system itself.

Adversarial Attacks

Threat: Using adversarial attacks, bad actors can manipulate input data to essentially trick the AI into making incorrect decisions. These attacks can be virtually imperceptible to humans, but can have devastating consequences. 

Danger: An adversarial attack can manipulate facial recognition systems to misidentify a person, or cause an autonomous vehicle to misidentify a stop sign.

Data Poisoning

Threat: Data poisoning occurs when an attacker purposefully injects incorrect or misleading data into the training data set that the AI is built on. This can cause the AI to output biased or inaccurate data.

Danger: Inaccurate data can wreak havoc on systems like health care and financial systems, potentially putting lives at stake. Data poisoning can also lead to long term instability in an AI system, making it less reliable over time. 

Model Theft

Threat: Attackers can potentially hack AI systems to steal or replicate their core models.

Danger: Model theft is a major security concern as it is possible for attackers to extract sensitive information from the model, in addition to creating a copy of the original model.

AI Intellectual Property Issues

A significant challenge also lies with AI intellectual property issues. Specifically, the question of who owns AI generated works has raised concern.

When AI systems are used to create outputs based on user prompts, they draw upon the data that they were trained on, and oftentimes they closely resemble them leading to copyright or trademark infringement risk.

And consider what happens when AI is used to write software.

Who owns the output - the organization that built and trained the AI, or the user who wrote the prompt?

Current patent and current copyright rules require originality from human input. But how much human intervention is enough? If you edit code generated by AI, is that copyrightable? The law is too new on this issue to know for sure. 

If AI generated code is not copyrightable, you may experience issues protecting your software product (if someone can get an actual copy to reproduce). 

In private transactions, you should ensure the proper ownership and license terms are in place to allocate ownership, versus license rights to all AI related IP including output and prompts. 

SVT can help you understand and properly manage these and other related issues core to the success of your business.

Prompt engineering can be used creatively to materially influence the quality of the resulting output, which makes ownership over output a significant gray area. 

The same issues can also be applied when AI is asked to create essays, novels, images, videos and other generative subjects.   

These are some of the legitimate questions around intellectual property infringement and AI that current laws do not adequately answer or that are potentially in flux. 

While they are unresolved, carefully consider who should bear the risk from potential infringement issues and the results you want regarding ownership of prompts and results from prompts–then ensure you have a lawyer to help you negotiate the right contractual language to achieve those results.

Conclusion

With AI there is tremendous potential for increasing productivity, efficiency, and innovation in business. But there is also tremendous risk for businesses who neglect to take AI data privacy, security, and intellectual property rights seriously. 

As a business that uses generative AI, it is imperative to have a solid understanding of the legal risks of artificial intelligence before adoption in your organization.

If you're considering adopting AI in your organization, SVTech Law Advisors can help you navigate the gray areas of artificial intelligence legal risks. For over 25 years, we've helped our clients with all aspects of technology law by delivering high quality solutions, focused on helping you meet your business goals. 

SVTech can help you draft contracts that thoughtfully allocate the risk of using AI between you and your counterparties and that define intellectual property ownership and license of AI's use, prompts and output. We can also advise you on adjacent topics like data privacy compliance, AI risk assessment and more.

Before you make moves to adopt AI in your organization, let SVTech guide you with comprehensive legal counsel. Contact SVTech Law Advisors today.