AI therapists are becoming a reality faster than the legal and ethical frameworks needed to regulate them can take shape. Who bears responsibility when an algorithm gives bad advice? How is the most intimate kind of data — mental health data — protected?

A Real Incident: When an Algorithm Caused Harm

In 2023, the chatbot Tessa — designed specifically to support people with eating disorders — was shut down after it was discovered to be giving users weight-loss tips and calorie-counting advice. For people with anorexia or bulimia, this is direct harm. The incident became a symbol of how critical safety is when developing AI for mental health.

10 Key Ethical Questions

Systematic reviews highlight several core themes: data privacy, informed consent, algorithmic bias along racial and gender lines, system transparency, the right to opt out of AI, and safety in crisis situations.

One especially pressing question: can AI respond appropriately to suicidal ideation? Research suggests that most current systems are not equipped to do so.

How the US Regulates AI Therapy

By 2025, the FDA had approved several digital therapeutics for mental health. The first, in 2017, was reSET — an app for substance use disorders. In March 2024, Rejoyn became the first FDA-approved app for treating depression, although its clinical trial did not demonstrate a statistically significant advantage over the control group.

No generative AI therapist has received approval yet. At a November 2025 hearing, the FDA stated plainly: "the metacognitive limitations of AI create significant risks, including potentially fatal consequences from incorrect information."

The EU AI Act: High Risk, High Requirements

Since August 2024, the European AI Act has been in effect. AI systems in healthcare are automatically classified as high-risk. This means mandatory conformity assessments, full technical documentation, continuous monitoring, and human oversight. Penalties for violations can reach 35 million euros or 7% of a company's annual global turnover.

What Users Should Know

If you use an AI app for mental health support, there are several things to keep in mind. The system should clearly disclose that you are interacting with AI, not a human. Your data — especially sensitive health data — must be stored in compliance with the law. In a crisis, the system must redirect you to a human professional or emergency services. If any of these safeguards are missing, you're looking at an unsafe product.

AI in psychotherapy is neither inherently good nor bad. It's a tool. Like any medical tool, it requires standards, oversight, and honesty with its users. Nearby is built on exactly these principles: transparency, safety, and respect for the boundaries of technology. Scientists and regulators are working on the rules — and we're keeping a close watch to stay on the user's side.