AI chatbots give misleading medical advice 50% of the time, study finds

A new study reveals that popular AI chatbots deliver medically inaccurate or incomplete information in half of all interactions, raising urgent safety concerns. | Contesto: cronaca

Punti chiave

• AI chatbots give misleading medical advice 50% of the time, study finds

Contesto

A landmark study published this week has found that leading artificial intelligence chatbots provide misleading or incomplete medical advice approximately 50% of the time. The research, which evaluated several widely-used public models, concluded that none of the systems produced a fully complete and accurate reference list in response to any medical prompt, despite often delivering answers with high confidence. The findings present a stark warning to the growing number of patients and healthcare professionals turning to these tools for quick medical information. Researchers systematically tested the chatbots with a range of complex medical queries, from symptom interpretation to treatment options for specific conditions. The consistent failure to cite correct, verifiable sources undermines the reliability of the information provided, creating a significant risk of misinterpretation. This "confidence gap"—where answers are presented authoritatively but lack factual rigor—is identified as a critical danger. Experts warn that a user, particularly one in distress or without medical training, is likely to trust a confidently stated answer, even if it is partially wrong or omits crucial caveats. The study's authors emphasize that this combination of inaccuracy and assured delivery could lead to real-world harm, including self-misdiagnosis, inappropriate treatment actions, or a dangerous delay in seeking professional care. The proliferation of AI chatbots in consumer technology has made them a first point of contact for millions seeking health information. Their 24/7 availability and conversational interface offer an appealing alternative to sifting through dense medical websites or waiting for a doctor's appointment. However, this very accessibility amplifies the risk, as the tools operate without the safeguards and ethical frameworks governing clinical professionals or accredited medical websites. Medical bodies and AI ethicists have long cautioned against using general-purpose chatbots for health advice, noting they are not diagnostic tools and are trained on broad internet data that can include biases and inaccuracies. This study provides concrete, quantitative...

Lettura DEO

Decisione di validazione: publish

Risk score: 0.1

Il testo è stato ricostruito dai dati editoriali disponibili senza aggiungere fatti non presenti nel record sorgente.

Indicatore di affidabilità

Verificata — Alta confidenza. Fonti affidabili confermano la notizia.

Il sistema a semaforo

Ogni articolo su DEO include un indicatore di affidabilità:

• 🟢 Verificata — Alta confidenza. Fonti affidabili confermano la notizia.
• 🟡 In evoluzione — Confidenza moderata. Alcuni dettagli potrebbero ancora cambiare.
• 🔴 Contestata — Bassa confidenza. Fonti in conflitto o incertezze rilevanti.

Questo sistema esiste perché chi legge merita di sapere non solo cosa è successo, ma anche quanto la notizia è solida.

Categoria: cronaca