Neural network error: how to deal with AI healers
In Moscow, a 68-year-old pensioner was hospitalized after she used a prescription from an artificial intelligence "medicine woman" to cleanse her intestines, media reported. If earlier pseudo-doctors and various healers gave questionable advice on their own behalf and had to answer to the law in case of harm to the health of their clients, now there are absolutely anonymous "healers" hiding under the guise of an AI avatar. Izvestia figured out how to deal with this phenomenon.
Decoction from the AI doctor
The situation with the 68-year-old Muscovite was described in one of the Telegram news channels. According to him, the woman subscribed to a certain "herbalist" on social networks, who published various video recipes for recovery. The pensioner decided to try a decoction with lemon and black pepper, which was supposed to help her clean her intestines. At night, the woman had a severe stomach ache, her blood pressure dropped sharply, she began to suffocate, and later bloody vomiting appeared. The Muscovite managed to call her family, who called an ambulance — and it saved her. As a result, the pensioner was hospitalized.
There are problems with neural network "healers" in other regions of Russia. However, such AI healers are already a truly massive phenomenon found in different countries of the world. In July, the Spanish media reported on dozens of profiles of doctors, healers, etc., which were created on one of the popular social networks using artificial intelligence. The videos they create get millions of views. Other media outlets have drawn attention to AI accounts that impose the use of various biological supplements.
Why are the elderly at risk?
Manipulation in the field of medicine, according to the head of the ST agency, Evgenia Sokolovskaya, is especially common.
— There have always been clickbait articles with loud headlines: "This product is killing you", "Gastritis can be cured in a week only with this remedy". These techniques have migrated to video content. And it is no coincidence that the image of a grandmother herbalist was chosen, which inspires confidence in authority and experience. The image of a doctor in a white coat or a man in a medical suit, the background of a hospital or a treatment room is also popular," she told Izvestia.
Such content often uses the phrases "doctors don't talk about it," "this prescription for treatment is from my grandmother." At the same time, such videos are often aimed at the elderly, who may "lack critical thinking skills in the field of AI development," the expert warned.
— After retirement, the loss of loved ones or friends, the elderly often remain with a limited social circle. Who else will cause a feeling of care, if not the person from the screen? The criticality to his proposals naturally decreases: so he listed all your symptoms on the list or described your day the way you would describe it yourself," Elizaveta Oleskina, director of the charity foundation "Old Age in Joy", explained to Izvestia.
The elderly, she clarified, were formed in an era when the print and television word had high authority, and the habit of trusting the "official" information channel is often transferred to modern formats — from newspaper ads to instant messenger videos. It also works the other way: someone remembers that "they lied a lot on TV, but on the Internet, finally, you can find out what was hidden from us." At the same time, some elderly people still have low digital literacy: it is difficult for them to verify the source and distinguish a fake medical center from a real one.
— Charlatans speak a language that is understandable and familiar to the elderly: "folk remedy", "tested by generations", "doctors hide". They turn to life experience ("they used to be treated with herbs and lived for a long time"), to distrust of official medicine, which some grandparents formed against the background of queues at clinics, burnt-out doctors or expensive private services. So now, when someone exploits the experience of these frustrations and resentments, it's not so important whether they use AI or not," the source said.
Problems with communicating with a real doctor push an elderly person into a space where an alternative is offered not by medical specialists, but by sellers, she added. Another risk factor is financial: the "miracle remedy" made from pepper and lemon can be served as an affordable alternative to expensive examinations and medications.
— The problem is not that the elderly "believe AI", but that the system does not offer anything more convincing and accessible. Therefore, it is not enough to prohibit content created using neural networks. It is necessary to develop a system of information support for the elderly in general," Elizaveta Oleskina emphasized. — In addition to teaching digital literacy, we need affordable medical counseling, as well as maintaining neighborhood and family ties.
Who earns money from AI doctors
According to the observations of experts of the Moshelovka platform, there are not individuals behind accounts with AI healers, but organized groups that work according to a conveyor model.
— The scheme looks like this: neural networks generate images of "herb grandmothers", "healers", "psychics", "elders" and "monks", then hundreds of accounts are created using anti‑detection browsers and virtual numbers. Content, responses to comments, communication scenarios — everything is done automatically. This is no longer a blog of one person, but full—fledged "farms" of synthetic characters," Alexandra Pozharskaya, deputy project manager of the Popular Front For Borrowers' Rights and the Moshelovka platform, explained to Izvestia.
At the same time, one person can now maintain several accounts with AI doctors at once, Evgenia Sokolovskaya added. The neural network helps at all stages — from the idea of a digital avatar to the final video.
"They usually create several characters, see which one fits the audience better, and then scale it up," she said.
Monetization can be different. The account is first promoted, and then they start selling a product or find advertisers. In some cases, the character is sold. At the same time, there is a demand for advertising on such blogs, and it is growing, the expert emphasized.
Who will be responsible for the harm from the AI medicine man
Each AI character has a specific person who systematically gives people medical advice, Konstantin Hansel, a lawyer at the Center for the Humanization of Law, drew attention to.
— If in this situation the fact of harming a person is confirmed, then law enforcement agencies can look for corpus delicti in the current legislation, — said the interlocutor of Izvestia. — Such a person can be charged under Article 235 of the Criminal Code of the Russian Federation, which provides for liability for illegal medical activity without a license. Other formulations may also be used.
Responsibility will depend on what exactly was advised to the person and what consequences it led to, added Yulia Ramzenkova, a member of the Russian Bar Association and lawyer at Leges Bureau. 118 ("Causing serious injury to health by negligence") or 109 of the Criminal Code of the Russian Federation ("Causing death by negligence") may be applied. 238 of the Criminal Code of the Russian Federation ("Production, storage, transportation or sale of goods and products, performance of works or provision of services that do not meet safety requirements").
"The main difficulty is to prove the connection between a specific recommendation and harm to health, as well as to establish who created and controlled such an account," the specialist noted. "There is no established judicial practice specifically on AI medicine men in Russia yet. There are cases involving illegal treatment and dangerous recommendations, but there is no separate practice on AI characters yet.
The difficulty may arise precisely when establishing the subject of responsibility: the investigation will need to determine the account owner, the source for the content generation, as well as the persons who made decisions about the publication of materials, Konstantin Hansel said.
In turn, Evgeny Kostromin, a lawyer and founder of the ContractCreation by Kostromin Legal boutique, said that it is possible to identify the owner of such accounts through law enforcement agencies at the request of the victim.
"The process includes official requests for data from platforms, banks, and telecom operators," he explained to Izvestia. — Traces usually remain: linked phone or email, payment details (donations, sale of courses or dietary supplements), advertising integrations and technical data.
In addition, law enforcement officers can study IP addresses, login history, and compare several pages based on technical and behavioral characteristics and find the same owner, said the head of the Department of digital Threat Analysis and Assessment at Softline Group (Infosecurity, part of Softline Solutions). Kristina Burenkova. And then we have to prove that a particular person managed the account, published recommendations, received income from it, and identify the link between the published advice and the consequences.
In Russia, social networks have responsibilities for monitoring certain categories of information and handling requests, said Svetlana Gazizova, owner of the UserGate AI security product. The law allows for judicial protection in cases of violation of rights due to non-fulfillment of these duties.
— But, of course, it is necessary to evaluate not the fact of content generation itself, but the content of the recommendation, the stated qualifications of the character and the possible consequences for the person. Instructions on self—medication and calls to refuse medical care should be considered especially carefully," the editorial interlocutor is convinced.
How to deal with AI healers
The problem with AI healers, according to Kristina Burenkova, is not in the neural network as such - a person may well write the same dangerous advice on their own.
— Rather, we need clear rules for cases where an artificially created character is posed as a real doctor, healer, or other specialist. Clear labeling requirements for such content and a mechanism for prompt removal after a well—founded complaint are important," she stressed.
It is also important to develop AI safety requirements, taking into account the consequences of its use, Svetlana Gazizova added. An entertaining AI character and a virtual "doctor" who recommends treatment need different levels of control.
"In the field of health, a mistake or deliberate manipulation can lead to physical harm, so we need a verifiable source, a responsible person and procedures for evaluating dangerous recommendations," explained the editorial interlocutor.
At the platform level, she proposed to introduce accelerated consideration of complaints about health threats, restriction of recommendations and monetization with a confirmed risk, as well as identification of related accounts.
— The legislation should also clarify the requirements for naming fictional specialists, disclosing the owners of commercial accounts about health and responding to dangerous advice. Responsibility should be determined by the actions of the people who create and distribute such content," the expert concluded.
Переведено сервисом «Яндекс Переводчик»