- New ChatGPT security guidelines are irritating customers
- Delicate subjects get rerouted to completely different AI fashions
- OpenAI has posted a response to consumer complaints
OpenAI is embroiled in one other controversy over AI mannequin switching in ChatGPT, with many paying customers livid that they are being rerouted away from their most well-liked mannequin at any time when dialog subjects get emotionally or legally delicate.
There are lots of threads on Reddit concerning the concern: in brief, ChatGPT launched new security guardrails this month, which take customers to a separate, extra conservative AI mannequin if the chatbot detects that it must be further cautious in its responses.
That is clearly annoyed and irritated plenty of customers, who wish to follow GPT-4o, GPT-5, or no matter they occur to be utilizing – particularly those that are paying. There’s presently no option to disable this conduct, and it isn’t notably clear when the switches occur.
“Adults deserve to decide on the mannequin that matches their workflow, context, and danger tolerance,” writes one consumer. “As an alternative we’re getting silent overrides, secret security routers and a mannequin picker that is now principally UI theater.”
Security routing
We’ve seen the sturdy reactions to 4o responses and wish to clarify what is going on.September 27, 2025
Sufficient of a fuss has been kicked up that OpenAI govt Nick Turley has weighed in on social media. Turley explains that the brand new security routing system is for “delicate and emotional subjects”, and works on a per-message and short-term foundation.
It is a part of a broader effort to enhance how ChatGPT responds to indicators of psychological and emotional misery, as OpenAI has beforehand defined in a weblog publish. From the consumer perspective although, the brand new guidelines are taking some getting used to.
OpenAI clearly has a accountability to take care of customers who may be susceptible and who may want further assist from an AI chatbot that is not fairly so expansive and freeform, and particularly younger individuals accessing ChatGPT.
For lots of customers venting their anger on-line although, it is like being compelled to look at TV with the parental controls locked in place, even when there aren’t any youngsters round. That is prone to be a problem we hear extra about within the coming days.
You may also like
