17/06/2026
This is not the article we planned to publish this week.
But sometimes reality interrupts the editorial calendar.
Our new piece comes from a situation we are currently experiencing inside a care-related project but the issue we discuss is much larger.
It is about what happens when mental-health AI is treated as a deliverable before it is treated as a responsibility.
It is about the legal and ethical danger of placing a chatbot inside a psychologically sensitive environment and pretending that the word “educational” removes the need for proper training, validation, human oversight, safeguarding logic, multilingual testing, and documented safety procedures.
It is also about the human cost of insisting on safety when the project around you would prefer speed and silence.
AI safety is not a delay.
In mental-health contexts, it is the work.
https://associationredefine.substack.com/p/before-the-chatbot-says-im-here-to-help-mental-health-ai-safety
A warning from inside a care project about mental-health AI, legal risk, safety validation, harassment, and what happens when care becomes inconvenient.