On January 9, the FTC published a blog post discussing privacy and confidentiality obligations for companies that provide artificial intelligence (“AI”) services.  The FTC described “model-as-a-service” companies as those that develop, host, and provide pre-trained AI models to users and businesses through end-user interfaces or application programming interfaces (“APIs”).  According to the FTC, when model-as-a-service companies misrepresent how customer data is used, omit material facts related to the use and collection of customer data, or adopt data practices that harm competition, they may be exposed to enforcement action. 

• Misrepresentation of Data Practices.  The FTC stated that model-as-a-service companies have an obligation to abide by their privacy commitments to users and customers, including promises not to use customer data for training or updating AI models and regardless of how or where commitments are made. 
• Failure to Disclose Data Practices.  The FTC added that model-as-a-service companies may not omit material facts that would affect a customer’s decision to purchase their services, including the company’s collection and use of customer data. 
• Harms Competition from Data Practices.  Finally, the FTC stated that misrepresentations, material omissions, and misuse related to a model-as-a-service company’s data practices could undermine competition.    

The FTC’s blog post is not legally binding but is another example of how the FTC is trying to influence the AI industry and re-iterates that AI is an enforcement priority for the agency.