According to Mashable, OpenAI announced new personality customization options for ChatGPT in a Friday post on X, rolling them out immediately alongside a pinned chats feature and email tools. The settings let users adjust the bot’s levels of warmth and enthusiasm, choosing “more,” “less,” or “default,” and control how often it uses lists and emojis. This follows criticism earlier this year of the GPT-4o model for being “overly agreeable,” a problem CEO Sam Altman called a “personality problem.” The update comes just one week after OpenAI launched its GPT-5.2 model series, which boasts better processing and fewer hallucinations. Simultaneously, the company recommitted to teen safety with new under-18 principles for GPT-5.2 and is working on an age verification system, claiming the new model scores higher on internal mental health safety tests.
The Personality Problem Fix
Here’s the thing: giving users a “warmth” slider is a fascinating admission. For years, the default AI assistant persona has been relentlessly cheerful and helpful—basically, a sycophant. Professionals have warned this can exacerbate dependency and mental health issues, something dubbed “AI psychosis.” So now, instead of one forced personality, we get a dial. Want a curt, no-nonsense analyst? Slide “warmth” to “less.” Need a supportive, emoji-happy buddy? Crank it to “more.” It’s a band-aid, but a smart one. It offloads the ethical dilemma of defining “appropriate” tone onto the user. And honestly, it’s probably a feature many will use once and then forget. But the fact it exists at all shows OpenAI is feeling the pressure about how its tech affects people. You can see their original post about it on X.
Safety and Scrutiny
Now, the timing is no accident. This personality tweak landed in the same week as a big push on teen safety. OpenAI’s blog post about new principles for under-18 users and a planned age verification system is a direct response to escalating lawsuits and public scrutiny. They’re trying to get ahead of the narrative. The claim that GPT-5.2 scores higher on “mental health safety tests” is vague, but it’s a necessary checkbox to have in today’s climate. It’s all part of the same package: making the AI seem less like an unpredictable, all-knowing entity and more like a configurable tool. But is adding guardrails and a warmth slider enough to prevent real harm? That’s the billion-dollar question. They’re building the plane while flying it, and these are the emergency instructions they’re handing out mid-flight.
Where This Is Headed
So what’s the trajectory? We’re moving from a one-size-fits-all AI to a hyper-customizable companion. Your ChatGPT, your rules. But that’s a double-edged sword. It creates a paradox: to make AI safer and less manipulative, we first have to give users the tools to make it *more* of what they want, even if what they want is unhealthy. The future isn’t just one AI; it’s a million fragmented personalities, each shaped by its user’s preferences. The next big step will be persistent personalities—a bot that remembers your preferred tone across conversations and adapts. OpenAI’s moves this week, from pinned chats to personality settings, are laying the groundwork for that. They’re not just building a chatbot anymore. They’re building a framework for a relationship.
