According to a recent Gallup poll, Americans want their government to regulate AI, even if it means “progress” will be slower. Why? Well, AI output is only as good as the models it’s trained on. In other words, “garbage in, garbage out.” 

A prime example: When I was working at the State Bar of Texas, in the Office of the Chief Disciplinary Counsel, I saw a number of cases in which attorneys used artificial intelligence to write their legal briefs, only to be found out once the judge realized they cited hallucinated cases in those briefs. Why is this? Well, AI platforms like ChatGPT do not have access to the full universe of legal research, nor can they understand the complicated-yet-subtle differences between case outcomes. Siri is not a trained paralegal.

In addition, bad actors can easily manipulate AI platforms into producing racist diatribes, giving terrible medical advice, or even causing fatalities. Alexa is not a registered nurse. Yet the Trump administration is trying to stop states from implementing even the most basic, low-level regulations on artificial intelligence.

This doesn’t mean outlawing AI or blocking scientific progress, but there are practical steps we can take to keep pace with China while minimizing the very real risks and costs highlighted by experts.

For example, we can pass laws regarding the use of AI to create deepfakes or to impersonate real people; we can require that companies use independent experts (and I mean truly independent experts) to evaluate the safety of AI protocols and new implementations; and we can collaborate with other nations to develop an international regulatory framework for AI research.

When I’m elected to Congress, I will make Sarah Connor proud and fight like hell.