Lately, content owners have targeted generative AI providers with infringement suits, which has made some consumers wary of potential infringement liability. Earlier this year, Valve, which operates Steam—the largest online game store—reportedly adopted a policy to reject any games with underlying AI-generated content due to infringement concerns.
Valve’s policy came to light after a developer posted on the “aigamedev” subreddit that Valve rejected his game submission, which included AI-generated content. The rejection notice cited AI-generated art assets that rely on copyrighted material owned by third parties. Since “the legal ownership of such AI-generated art is unclear,” Valve could not ship the game.
A Valve spokesperson later noted that their goal is not to discourage the use of AI, but rather, that Valve’s “review process is a reflection of current copyright law and policies, not an added layer of [Valve’s] opinion.” The “reflections” giving Valve pause are likely the rise of lawsuits against providers of generative AI products. Pending lawsuits around the county are testing the questions of whether the use of copyrighted content to train an AI model constitutes copyright infringement and whether (and to what extent) content generated by generative AI is or can be considered infringing.1
In response to the infringement fears, some generative AI providers are trying to get in front of the issue by offering to protect their customers from claims through indemnification.2By way of background, indemnification is a contractual commitment by a party to a contract to compensate another party to the contract from certain defined losses that they incur. Indemnification is often also paired with an obligation to defend another party from the claims that give rise to the losses. In the software industry, software providers often indemnify and defend their customers from third-party infringement claims brought against their customers that arise from the customer’s use of the provider’s software, on the general principal that the risk of infringement claims is foreseeable, and the software vendor is most likely to be responsible for the claim and best positioned to defend the suit and address liability.
Many generative AI providers are hesitant to provide this level of indemnity for their AI systems or the output of their AI systems, however, for many reasons. Three particular reasons stand out:
- There are risks around the training data. Most generative AI tools were trained using large amounts of “scraped” data, which has led content owners to bring copyright infringement suits alleging that the scraping of their data and use of their content to train the generative AI model was an infringement of their copyright.
- Generative AI tools are responsive to user requests and give users the ability to control outputs through prompting, including by requested infringing content. While generative AI providers can and have implemented technical measures such as filters in an attempt to screen infringing content, there is still a degree of risk when a customer specifically requests content that may be infringing.
- Unlike traditional software tools, which are built to perform a specific task, generative AI tools are general-purpose tools that can be and are used in unforeseeable ways. AI providers may be hesitant to agree to protect users from totally unforeseen uses that the AI provider did not contemplate in developing their products.
Yet, despite these risks, providers of generative AI tools are increasingly announcing commitments to indemnify users from infringement claims. For example, Microsoft announced its Copilot Copyright Commitment, with customer concerns for copyright infringement in mind. Effectively, the program expands upon indemnification coverage for its customers, subject to several exclusions, including intentional misuse to generate harmful content.
Microsoft outlined exactly why it was offering the program in three points. First, Microsoft wants to stand behind its customers when they use their products. Second, Microsoft is the better party to bear any losses from a potential infringement. And finally, Microsoft has already implemented strong guardrails into their Copilots that respect copyright, which reduces the likelihood of infringing content anyway. Microsoft believes the world “needs AI to advance the spread of knowledge and help solve major societal challenges,” so the Copilot Copyright Commitment survived any risk-balancing analysis. Additional companies, including OpenAI, Anthropic, Google, Adobe, and Getty Images, have each come out with similar announcements for their generative AI tools. These announcements suggest that AI vendors are trying to address the concerns of enterprise customers by taking matters into their own hands and proactively offering protection through indemnity.
However, both the type of defined indemnification commitment and the extent of the offered indemnification prove that this is still a sensitive topic. It is thus important to retain qualified attorneys to determine the company’s circumstances for using AI and which approach is best suited to avoid potential areas of infringement.
Regardless of whether a generative AI company has decided to play it safe or indemnify their customers, there are liabilities when employing AI that will need to be mitigated through proper policies and procedures. Each company will vary in its approach to indemnification and to managing the liabilities that come with AI depending on its intended use.
The emergence of AI has added a new dimension to copyright that the law is still forming. While companies continue to experiment with its capabilities, attorneys are best suited to guide and advise on the development of strong guardrails for AI’s intended use at the company.
[1] A number of providers of AI-generated images are involved in lawsuits regarding copyright and AI. See Ellen Sheng, In Generative AI Legal Wild West, the Courtroom Battles are Just Getting Started, CNBC (Apr. 3, 2023), https://www.cnbc.com/2023/04/03/in-generative-ai-legal-wild-west-lawsuits-are-just-getting-started.html.
[2] See Neal Suggs & Phil Venables, Shared Fate: Protecting Customers with Generative AI Indemnification, Google (Oct. 12, 2023), https://cloud.google.com/blog/products/ai-machine-learning/protecting-customers-with-generative-ai-indemnification and Brad Smith & Hossein Nowbar, Microsoft Announces New Copilot Copyright Commitment for Customers, Microsoft (Sept. 7, 2023), https://blogs.microsoft.com/on-the-issues/2023/09/07/copilot-copyright-commitment-ai-legal-concerns/.
A number of other providers have also joined in offering users indemnification. See Jon Jekel, OpenAI Joins Other Generative AI Companies in Offering Indemnity for Users Against (Some) Third-Party Infringement Claims, JDSupra (Nov. 17, 2023), https://www.jdsupra.com/legalnews/openai-joins-other-generative-ai-1871431/.
