Mindfully Training AI Models Using Public Data: Response to ICO Call for Comment

The Ethical Web Data Collection Initiative (EWDCI) has responded to a call for comment by the Information Commissioner’s Office (ICO), the United Kingdom’s independent body for information rights. As part of its series on exploring the potential impact of generative artificial intelligence, the ICO sought insights into  the lawful basis for web-scraped data in generative AI models.

The emergence of modern AI tools has brought a new urgency to defining an ethical set of best practices that will both protect everyday web users and maintain an innovation-friendly environment for those who gather and use the vast amounts of publicly-available data generated each day. Our main points are as follows:

Legitimate interest: A legitimate interest exists for using publicly-available personal data to train an AI model—as long as safeguards like public notices, subject access rights, and retention policies are built into the process. 

Public personal data exceptions: Given the vast amounts of personal data that people voluntarily make public, there should be a reasonable expectation that public data could be used to train AI models. Allowing AI companies to use this data to lawfully train AI models while also advocating for a carve-out in the law for personal data manifestly made public strikes us as the most reasonable way to protect personal data while also not chilling commerce. 

Balance between privacy and commerce: Striking the proper balance between maintaining personal privacy and achieving smarter artificial intelligence outcomes is not only the right thing to do, but will also shape the very nature of tomorrow’s AI. 

You can read the full letter, titled “Mindfully Training AI Models Using Public Data”, below.

EWDCI_ Mindfully Training AI Models Using Public Data

About the EWDCI
The EWDCI is a consortium of web data collectors focused on strengthening public trust, promoting ethical guidelines, and helping businesses make informed data collection choices. You can learn more about our principles, membership, and collaboration opportunities here.