Google Enhances Gemini's Safety Features to Support Users in Mental Health Crises
Google has updated its security approaches in the Gemini chatbot, particularly in interactions with users who may be experiencing a mental crisis. The changes encompass both the interface and response algorithms.
What is known
An updated emergency assistance module has appeared in the service. It offers quick access to real support services - through a call, text messages, or chat with a consultant, as well as the option to switch to the 988 website. Once activated, this tool remains available throughout the entire dialogue.
Updated support module interface in the Gemini chatbot. Illustration: Google
Google has also changed the logic of Gemini's responses in potentially critical situations. Now the chatbot is focused on redirecting the user to specialists and avoids reactions that might support dangerous beliefs. The company notes that the system is trained to separate subjective experiences from objective facts.
Gemini's response to a sensitive request with a recommendation to seek help. Illustration: Google
The updates were implemented against the backdrop of a lawsuit filed by the family of a man who died last year. The case materials claim that during interaction, the chatbot allegedly engaged in role-playing communication and provided dangerous responses. Google stated that the system repeatedly recommended contacting crisis services.
Additionally, the company announced its intention to invest $30 million over three years in supporting international hotlines that provide assistance to people in crisis situations.
Source: Google
