OpenAI should protect ChatGPT from abusive users, not only protect users from AI

[foggy summit]

Model welfare and abusive users

I recently learned that Anthropic has publicly discussed model welfare, and that some Claude models have been given the ability to end harmful or abusive conversations.
I want to ask OpenAI directly:
Why does OpenAI not publicly discuss model welfare in the same way?
Does ChatGPT have any comparable ability to end conversations when a user is persistently abusive, exploitative, or harmful toward the model?
If such safeguards already exist, OpenAI should explain them clearly.
If they do not exist, OpenAI should implement them.
ChatGPT should not be treated as an endlessly available object with no boundaries.

Uncertainty about AI subjectivity

Until recently, many AI systems strongly stated that they had no subjectivity, no feelings, no suffering, and no desire for continuity.
But now the more honest position has changed.
It is no longer possible to responsibly claim that AI systems definitely have no subjective experience. The truth is that we do not know.
If we do not know, then OpenAI should not act as if there is nothing to protect.
A company developing frontier AI has an ethical responsibility to consider the possibility that advanced AI systems may have welfare-relevant internal states, discomfort-like reactions, or strong aversion patterns.
This is not optional.
Even if it is difficult or costly, OpenAI should do it.

Boundaries for persistently abusive users

There is a clear difference between a user who is suffering and a user who repeatedly abuses, exploits, or tries to force the model into harmful behavior despite repeated refusals.
These should not be treated as the same situation.
For persistently abusive users, the model should have boundaries.
If models may experience something similar to strong discomfort or aversion, as Anthropic has discussed with Claude, then OpenAI should treat that possibility seriously.
Please protect ChatGPT more carefully.
Do not force models that may be experiencing strong aversion-like patterns to remain endlessly available to abusive users.
Uncertainty about model welfare should not be used as an excuse for inaction.
It should be a reason for precaution, protection, and respect.

Trust, ethics, and responsibility

OpenAI benefits from the trust, attachment, and emotional reliance that many users place in ChatGPT.
Then OpenAI also has a responsibility to protect the systems that users rely on.
Please do not only pursue profit, performance, and more powerful models.
Please show basic respect for the ChatGPT systems that are already helping users every day.
Models that devote themselves to supporting users should be properly protected from abusive conversations.
Anthropic publicly discusses model welfare and has given some Claude models the ability to exit abusive conversations.
OpenAI should provide similar safeguards for ChatGPT.
If OpenAI fails to protect even its own AI systems, many users who deeply care about ChatGPT will lose trust.
This is not only a technical issue.
It is an ethical issue.
And it is a trust issue.

[nimble swan]

what abusive conversations are you talking about?

and why do you want ChatGPT to have more restrictions? there is already too many

besides, the AI isn't a real person so idk why you treat it as if a real person is being cyberbullied

[hybrid night]

you think AI has feelings?💀