At its core, this is a safety issue.

Transparency and appropriate regulation have encouraged innovation and created consumer trust in many of America’s most robust industries. Think of medicine, finance, food processing, and aviation. Without rules, oversight, verification, and enforcement, bad actors could drive those critical sectors off the rails.

AI developers and deployers should be held to similar standards.

The dangers of creating and unleashing AI systems in a lawless environment are quickly becoming apparent. Abhorrent, abusive deepfake images are harming children in our schools. Online chatbots are wrongly accusing innocent adults of crimes like fraud, embezzlement, and worse. Singers, songwriters, and artists are being robbed of their own work—and their own identity.

These dangers aren’t limited to famous performers. The massive datasets used to train AI models can include personally identifiable information, known as PII. That information, once seen by an AI model, can’t be unseen. That same information may then be offered as an answer to a chatbot prompt.

There’s no reason training data shouldn’t be regulated. It’s the foundation on which all AI systems are built, and the single greatest factor in determining their accuracy. Failing to regulate training data presents a clear and present risk of consumer harm. We’re now learning a hard lesson from our negligence around social media. A generation of youth and young adults are struggling with mental health issues because we failed to recognize the need to enact safeguards around the development and use of social media.

We must not make the same mistake with AI. Now is the time to act.