Securely Collaborate on AI with AWS Clean Rooms ML: Unlocking New Opportunities for Companies

Amazon has introduced a groundbreaking privacy-preserving service in preview, enabling AWS customers to deploy “lookalike” AI models tailored for unique collaborations between companies. This innovative offering, named Clean Rooms ML, is an extension of AWS’s existing Clean Rooms product. It eliminates the necessity for AWS customers to share sensitive proprietary data with external partners, paving the way for the development, training, and deployment of AI models.

“With Clean Rooms ML, you can train a private lookalike model using your combined data,” stated Swami Sivasubramanian, VP for Data and Machine Learning Services at AWS, during a keynote at AWS re:Invent this morning. “You maintain control over your models and can delete them when they are no longer needed.”

Clean Rooms ML empowers customers to use a limited sample of customer records to create a broader set of similar records with a partner—essentially forming a lookalike AI model. For instance, an airline could extract insights about its loyal customers and collaborate with an online booking platform to extend promotions to new users who exhibit similar traits.

Additionally, Clean Rooms ML provides adjustable controls to tailor model outputs to meet specific business objectives, as noted by Sivasubramanian. Looking ahead, AWS plans to introduce features specifically for healthcare applications.

In a related development, Amazon also launched Clean Rooms Differential Privacy, a fully managed service under the Clean Rooms umbrella. This service ensures customer data remains anonymized, generating “aggregate insights” that allow businesses to gather combined analyses on advertising campaigns, investment strategies, and clinical research without disclosing proprietary information.