AI and mental health, Google intervenes on Gemini’s limits

The heart of the update lies in a new interface defined as “one-touch”, designed to simplify immediate contact with professional assistance services. When the system detects signs of crisis or self-harm intentions, it activates a help form that remains visible and accessible for the entire duration of the session. In addition to the technical component, the restyling includes responses developed with a more empathetic tone, designed in collaboration with clinical experts to encourage the use of traditional therapeutic paths.

Alongside the software evolution, the company has formalized a $30 million investment over three years into emergency hotlines on a global scale. While acknowledging that AI cannot replace medical or psychiatric advice, Google implicitly admits that a growing number of people are using chatbots as their first point of contact for sensitive health matters. This awareness is pushing for a review of the sector’s safeguards, often criticized for failing in the past to prevent the machines from providing dangerous indications of eating disorders or planning of violent acts.

While independent analyzes often place Mountain View solutions a step above the competition in terms of security, the risk of systemic failure remains real. Like OpenAI and Anthropic, Google is trying to lock down its linguistic models to prevent freedom of interaction from resulting in a danger to public safety. The current effort aims to transform the digital assistant from a potential element of instability to a proactive filter for risk prevention.