The Top 5 Common Issues to Look out for in AI Technology Agreements
As companies have been adding AI services into their workflow, naturally, there comes an additional set of technology agreements with AI service providers. As with any new technology, there are contractual issues that everyone needs to be made aware of. We’ve compiled five of the most common issues we see in AI Technology Agreements that our clients should look out for.
#1: Representation and Warranties for AI Output Quality.
What it means: When your company uses a provider’s AI software, you’ll want to understand how the provider measures the quality of its output – and how it will remedy any shortfall. If the output is riddled with mistakes or has severe lag time, it’s critical to know whose responsibility it is to fix it and how fast it will be done. Otherwise, you’ll just be paying for an inefficient service. You will also want to make sure the AI is compliant with applicable law.
What to look for:
Existing language regarding the AI Service Provider’s representations and warranties surrounding the legality, accuracy and quality of the AI model and output;
Lack of language regarding the AI Service Provider’s representations and warranties surrounding the legality, accuracy and quality of the AI model and output; or
Inclusion of language regarding the AI Service Provider’s representations and warranties surrounding the legality, accuracy and quality of the AI model and output.
#2: Ownership Rights for AI Output
What it means: When your company engages a third-party Service Provider, you typically have the express intent of ensuring you can use the outputs forevermore, and for any reason. It’s important the Agreement reflects ownership rights accurately; you don’t have to face a claim from other clients of the Service Provider. Typically, AI Service Providers claim AI output can’t be owned by those using the service, because outputs can be similar across clients. While this may be true, the input leading to an output varies across the board and while the output may be similar, since it is derived from your data, it should be yours to use forevermore.
What to look for:
Lack of language stating Service Recipient shall own the output from its AI usage; or
Inclusion of language stating Service Recipient shall own the output from its AI usage.
#3: Use of Service Recipient’s Data
What it means: Granting the Service Provider unfettered use of your data could prove to be beneficial to competitors who may also be using the service. It could also lead to your data being used for matters exceeding the scope of the agreed upon engagement.
What to look for:
Existing language granting the AI provider the right or license to use Service Recipient’s data to broadly train its AI models;
Existing language granting the AI provider the right or license to use Service Recipient’s data other than just to provide the services; or
Inclusion of an obligation for Service Provider to use Service Recipient’s data solely for providing the services, and if training its AI models, only training for Service Recipient’s use.
#4: Indemnity for AI Outputs
What it means: If for whatever reason, the output of the AI model is subject to a claim for intellectual property infringement, you would need protection – ideally from the AI Service Provider.
What to look for:
Lack of an obligation for Service Provider to indemnify Service Recipient for intellectual property claims arising from its output (including content); or
Inclusion of an obligation for Service Provider to indemnify Service Recipient for intellectual property claims arising from its output (including content).
#5: Documentation of AI Provider/Practices
What it means: Often, Service Providers outsource their AI services to companies who own and operate AI models. Sometimes, the Service Provider has such capacity itself. If the former, it’s important to have details of who their provider is and where they can be contacted for your own data protection purposes (i.e. audits, data breach or incident, etc.). If the latter, then for the same reasons it’s important to ensure the Service Provider is willing to provide information regarding its methods and practices to the extent needed.
What to look for:
Existing language regarding the Service Provider’s AI provider;
Lack of specificity regarding the Service Provider’s AI provider; or
Inclusion of language regarding the Service Provider’s AI provider and how to request information regarding its methods and practices to the extent needed.
While not exhaustive, these five items are the most common concerns stemming from AI Technology agreements. If your company is considering AI technology use, and are working through similar AI Technology agreements, the Parsus team can help you navigate these contracts.