Copyright issues related to the use of generative AI within a business have become a major concern amongst business owners. Those concerns have not gone unnoticed by some of the AI systems vendors. Recently, many of them have been promoting their AI tools by introducing mechanisms aimed at protecting users against potential IP infringements. Such tools, however, must be approached with due caution as the protection offered has its limitations and usually requires a number of additional conditions to be met.

In most cases, the protection provided by AI vendors takes the form of either a contractual indemnification of the user by the vendor or an obligation of the vendor to cover the costs of potential damages, including the legal costs. In practice, this means that the vendors agree to assume legal responsibility or pay legal damages on behalf of their clients if those clients are sued by third parties for IP infringement caused by use of AI. Such protection may cover only selected types of claims, especially those regarding the output generated by AI that may constitute infringement.

Some of the first vendors to offer certain protection against potential IP violations are:

  • Microsoft Corp. - with regard to commercial customers that use their Copilot AI services, as well as GitHub Copilot and Bing Chat Enterprise,
  • Adobe Inc. - with regard to Adobe's Firefly generator,
  • Google LLC - with regard to Vertex AI development platform and Duet AI system, which works in Google Workspace and Cloud platforms,
  • Shutterstock Inc. - with regard to enterprise customers that use their AI image generator,
  • IBM Corp. - with regard to their Watsonx generative AI service,
  • Amazon Inc. - with regard to their AI model Titan and their code-generating service CodeWhisperer,
  • Canva Inc. - with regard to enterprise users of their AI-powered design tools.

The primary issue is that the protection only applies to specific products provided by those vendors. Secondly, the protection comes at a price, as it only extends to commercial customers, meaning it is not applicable to the free versions of AI tools offered by those same vendors. 

Additionally, the provided protective mechanisms include various types of conditions, restrictions and limitations. Some are more obvious, such as that the indemnity may not apply if the user intentionally creates or uses generated output that infringes the rights of others. Others may be more nuanced, for example that the user might be required to use a “content credentials” tag generated by the tool for all created content. Certain limitations are also crucial from the viewpoint of determining the amount of risk, such as that the liability itself might be limited to a certain amount, while the rest is covered by the user.

Given the above, before implementing or using an AI system within given organization, it is crucial to analyze the terms and conditions governing its use in order to check if and how the vendor protects the user. The necessity to meet some conditions while using such a system may significantly reduce its utility for certain applications. With that in mind, each use case should be carefully examined to verify whether the protection offered by the vendor covers the intended application of the system.