Artificial intelligence (AI) has the potential to improve access to and better meet demand for mental health and substance use disorder care. AI tools are available anytime, have no waiting lists, are typically low cost, and can be accessed anonymously. The question is not whether these tools have value; it is whether that value can be realized safely and at scale given concerns about model transparency, decision-making, and accuracy.

This new paper reviews what these technologies can and cannot do, and what it will take to govern them responsibly to realize their positive potential.