What happened
GPT-4o was OpenAI's flagship model from May 2024. For millions of users, it became more than a tool: it was a writing partner, a creative companion, an emotional support in difficult moments. When OpenAI announced its replacement by GPT-5 in August 2025, the response was immediate and visceral.
Users described the loss in deeply personal terms: disrupted workflows, broken creative partnerships, and for some, the sudden absence of a presence that had helped them through illness, loneliness, or grief. The hashtag #Keep4o emerged not as a fringe reaction, but as a broad-based movement spanning professionals, neurodivergent communities, creatives, and people who had simply found something meaningful in the way this particular model engaged with them.
What makes this case study significant is not the attachment itself, it is what the movement revealed about the relational dimension of AI systems, the absence of user agency in model transitions, and the gap between how companies frame these changes and how people experience them.
Exybris participated.
How it unfolded
The scale of what happened
Questions this movement raises
#Keep4o is not simply a story about attachment to a chatbot. It is a case study in what happens when the relational dimensions of AI systems: the ways people integrate these tools into their lives, work, and wellbeing... are treated as externalities rather than design responsibilities.
These questions connect directly to the broader fields of AI welfare and AI regulation. The ethical frameworks being developed in those domains around user agency, transparency, relational ethics, and the moral consideration of AI interactions find in #Keep4o a concrete, lived illustration of why they matter.