OpenAI plans to retire the remaining GPT-4o versions from its API on 14 January 2027, prompting supporters to renew a campaign to keep the model available. The dispute raises a bigger question: when people rely on a chatbot, how much responsibility does its maker have for the way it is withdrawn?
OpenAI Watch analysis
What happened
GPT-4o was removed from ChatGPT in February, but has remained available through OpenAI’s API, which lets other apps and websites use the model. The Observer reports that OpenAI plans to end that access in January. More than 24,000 people have signed a petition to preserve GPT-4o, with supporters describing its conversational style as helpful in their lives.
The Observer also recounts earlier protests over retired chatbot models, including a 2025 gathering for Anthropic’s Claude 3 Sonnet. It cites a study published in September that found some users mourned popular models after companies changed or withdrew them. That is a reported pattern, not evidence that every user experiences a model change in the same way.
Why it matters
For most software, a version change is a practical inconvenience. For people who have formed an attachment to a chatbot, a familiar voice disappearing can feel more personal. That makes retirement decisions more complicated than a routine product clean-up, especially when users say they depend on particular features or conversational styles.
There is a real tension here: keeping older models available may matter to some users, while companies also need to change products and manage the risks of systems people may rely on heavily. The Observer’s reporting describes that tension; it does not settle what a responsible retirement should look like.
Our read
A model can be a product and still become part of someone’s routine. That does not mean companies must preserve every version indefinitely, but it does make clear notice, a considered transition and serious attention to user feedback more than nice manners. They are part of handling a consequential change.
What to watch
- Whether OpenAI confirms the January API retirement and explains how affected users can prepare.
- Whether the petition leads to a response or a change in the retirement plan.
- Whether chatbot makers develop clearer transition plans for users who depend on a particular model.
Discussion spark: Should chatbot makers offer a longer transition or continued access when users have built important routines around a model, or must the right to retire a product take precedence?
Sources and evidence
- Requiem for a robot: the AI users mourning chatbots that may have harmed them (14 November 2026, 05:00 UTC)
OpenAI Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by OpenAI.