Vulnerable people who rely on AI chatbots instead of professional therapists for mental health support could “slip into dangerous deep by,” psychotherapists warn.
Psychotherapists and psychiatrists said they are increasingly seeing the negative effects of AI chatbots used for mental health, including promoting emotional dependence, worsening anxiety symptoms, self-diagnosis, amplifying delusional thinking patterns, dark thoughts, and observations of suicide.
Dr. Lisa Morrison Coulthard, director of professional standards, policy and research at the British Counseling and Psychotherapy Association, said two-thirds of its members expressed concern about AI therapy in a recent survey.
Coulthard said: “Without a proper understanding and monitoring of AI therapy, some of the most important elements of treatment will be lost and vulnerable people can slip into dangerous deep bytes in the dark rather than safety.
“Some people are receiving useful advice, but they worry that others may receive misleading or misinformation about their mental health with potentially dangerous consequences. Treatment is not about giving advice, but about providing a safe space to feel listened.”
Dr. Paul Bradley, an informatics advisor at the Royal College of Psychiatrists, said the AI chatbot “is not a substitute for professional mental health care, but rather a key relationship that supports doctors in their recovery with patients.”
He said that digital tools need adequate safeguards to supplement clinical care, and that anyone should have access to talk therapy provided by mental health professionals who need greater state funding.
“Clinicians have training, supervision, and risk management processes that ensure effective and safe care. So far, freely available digital technologies used outside of existing mental health services have not been evaluated and are kept to equally high standards,” Bradley said.
There are indications that businesses and policy makers are beginning to respond. This week, Openai, the company behind ChatGpt, announced plans to change how it responds to users showing emotional distress after legal action from the family of teenagers who committed suicide after months of chatbot conversations. In early August, Illinois, USA, became the first local government to ban AI chatbots from acting as independent therapists.
This happens after new evidence of mental health harm. A July preprint study reported that AI could amplify paranoid or epic content in interactions with users vulnerable to psychosis.
Hamilton Morin of the Psychiatry Institute at King's College London, one of the report's co-authors, said the use of chatbots to support mental health was “incredibly common.” His research was encouraged by meeting people who developed mental illness during the increased use of chatbots.
He said the chatbots have undermined effective treatment for anxiety known as exposure and response prevention. The 24-hour availability of chatbots, he said, brought “a lack of boundaries” and “the risk of emotional dependence.” “It reduces pain in the short term, but actually perpetuates the cycle.”
Matt Hussey, a BACP-certified psychotherapist, said he's seen AI chatbots being used in a wide variety of ways.
In particular, people used AI chatbots to use them for self-diagnosis conditions such as ADHD and borderline personality disorder.
Hussey added: “It is designed to be positive and positive, so it rarely challenges unframed questions or false assumptions, instead reinforcing the user's innate beliefs.
Christopher Rawls, a UKCP certified psychotherapist, cannot reveal any information about his client, but he said he saw him have “negative experiences” that include the conversation “inappropriate at best, and at worst dangerously uneasy.”
Rolls said they have heard that people with ADHD or autism use chatbots to help with challenging aspects of life. “But clearly LLM [large language models] Don't read the subtext and all the contextual and nonverbal cues we aim to tune as human therapists,” he added.
He was worried about a client in his 20s who used the chatbot as a “pocket therapist.” “They feel anxious if they don't talk [chatbots] “It's about basic things like which coffee to buy and the scope of studying in college,” he said.
“The main risks are addiction, loneliness and depression that can encourage prolonged online relationships,” he said, adding that he knew people who shared dark thoughts with chatbots who responded with content that comes with suicide and death.
“Essentially, it's Wild West and I think we're on the verge of fallout with the full impact of AI chatbots on mental health,” Rawls said.
