Montreal, October 10, 2025 - On the occasion of World Mental Health Day, Mila is proud to announce the launch of the AI Safety Studio and its first initiative to mitigate the risks associated with the misuse of chatbots by users with psychological distress.
This project will focus particularly on mental health issues related to the use of chatbots by young people by developing concrete safeguards to detect and block harmful content, including suicide assistance filters.
The goal for the next year is to launch AI guardrails to detect potentially dangerous interactions with conversational agents and prevent harmful responses, redirecting the user to appropriate public health assistance channels.
“AI can be an extraordinary tool, but without proper safeguards, it can also cause harm. The Mila AI Safety Studio is designed to help protect users and build a framework of trust around this powerful technology to enable its responsible adoption. To develop a complete and effective solution, the initiative will require close collaboration beyond the technical realm, involving legislators, educators, health professionals, and other key sectors, since AI safety is as much a social and ethical challenge as it is a technological one”, stated Simona Gandrabur, AI Safety Studio Lead at Mila.
Accelerating Responsible Adoption of Trustworthy AI
Over the coming months, the Studio will develop:
- Intelligent filters to block AI-generated content that assists or encourages self-harm or suicide
- Reliability tests to evaluate the safety and robustness of conversational and generative AI models
- Psychological and ethical risk-assessment tools in collaboration with mental health, education, and social science experts
- Compliance frameworks to guide the responsible use of AI across public and private sectors.
Three main priorities will guide the AI Safety Studio's work:
- Guardrails: Developing independent, external control mechanisms to prevent harmful interactions with Large Language Model applications.
- Multi-turn Benchmarking: Creating frameworks to rigorously test AI safety in conversational and agentic contexts.
- Alignment: Conducting research to improve LLM alignment with human values.
The Studio is committed to transforming academic research into tangible protections for young people while fostering pragmatic technological development and promoting educational and legislative collaboration.
This project is carried out by Mila through the targeted use of funds from the Canadian Institute for Safe AI. It represents a responsible and transparent use of public resources to strengthen the safety and reliability of AI technologies, marking a foundational step toward a sustainable ecosystem of trust that benefits citizens, institutions, and the Canadian economy.
