Responsibly Designed AI Certified Products

Digital Promise has verified that the following products have met the certification criteria and earned the Responsibly Designed AI product certification. These AI-powered edtech products submitted applications that underwent a rigorous assessment process to certify that the edtech developers designed the AI in the product to reduce algorithmic bias, provide transparency about data collection and security, include labeling of generative AI content, and equip educators with agency while engaging with AI outputs.

To earn the Responsibly Designed AI product certification, product teams must submit evidence to demonstrate the following:

1. The applicant has a public, free-to-access privacy policy, or similar document, to help current and prospective users understand:
   • What nonpublic personal information (NPI), including PII, data is collected, how it is stored, and for how long it is stored before being destroyed;
   • How this data is or may be used (e.g. to train the AI model);
   • Whether this data is shared with or sold to third parties; and
   • What control the user has over this data.
2. The applicant has a public, free-to-access policy or other document describing how the product ensures data security, including a data breach or other incident response plan, as well as states which relevant data privacy laws this policy complies with (e.g. FERPA, COPPA, GDPR, etc.).
3. The applicant has documentation about how the product team has sought to acknowledge and reduce bias in the AI model from development through implementation. The applicant has provided two examples of instances where these monitoring processes identified bias and described the changes made to mitigate the bias. Additionally, users can report instances of bias through the product directly.
4. There are labels within the product to indicate when AI has been used in content generation that a user might mistake as being created or shared by a human (e.g. chat communication, images, word problems, literature, narrative feedback, etc.).
5. Educators or other individual users can override at least one AI decision or recommendation made to support learning.