Welcome to the forefront of conversational AI as we explore the fascinating world of AI chatbots in our dedicated blog series. Discover the latest advancements, applications, and strategies that propel the evolution of chatbot technology. From enhancing customer interactions to streamlining business processes, these articles delve into the innovative ways artificial intelligence is shaping the landscape of automated conversational agents. Whether you’re a business owner, developer, or simply intrigued by the future of interactive technology, join us on this journey to unravel the transformative power and endless possibilities of AI chatbots.
New “Warmth” and “Enthusiasm” sliders allow users to fine-tune the chatbot’s tone beyond static presets.
Addressing persistent user complaints about its artificial intelligence (AI) being either too robotic or overly sycophantic, OpenAI has introduced new granular controls for ChatGPT’s personality. Rolled out Friday, the new “Warmth” and “Enthusiasm” sliders allow users to fine-tune the chatbot’s tone beyond static presets.
Alongside tone adjustments, the update adds toggles for emoji usage and formatting density. While these features offer more personalization, the company recently decided to postpone its controversial “Adult Mode” (which would permit Not Safe For Work content) until 2026.
OpenAI has deployed four specific sliders: Warmth, Enthusiasm, Emoji, and Headers & Lists. Users can toggle each setting between three distinct levels: “More,” “Less,” or “Default.”
You can now adjust specific characteristics in ChatGPT, like warmth, enthusiasm, and emoji use.
Now available in your “Personalization” settings. pic.twitter.com/7WSkOQVTKU
– OpenAI (@OpenAI) December 19, 2025
Augmenting the broader “Personality Presets” (Professional, Candid, Quirky) introduced with GPT-5.1 in November, these granular controls offer more precision. Unlike the presets, which apply a blanket style, the new sliders allow for mix-and-match customization.
A local-first browser extension that centralizes your prompts and chains across ChatGPT, Claude, Gemini, Grok, AI Studio, and Mistral – with a free 60-item plan.
Organize: Prompts in folders with tags, favorites, and fast search.
Automate: Run multi-step chains in Auto, Manual, or Paste mode across all major AI chats.
Control: Keep everything private: data stays in your browser, with import/export and version history.
For example, a user could configure the model to display high warmth but low emoji usage.
Directly responding to conflicting user feedback regarding GPT-5.2, the update addresses complaints that the model felt “cold” or “robotic.” Conversely, earlier models faced criticism for “sycophancy”, the tendency of AI to agree with users regardless of facts to appear helpful.
A recent study on AI sycophancy published in Nature quantified this issue. According to the data, GPT-5 displayed a 29% sycophancy rate in interactions, compared to 70% in competitor DeepSeek-V3.1.
OpenAI CEO Sam Altman has previously acknowledged the difficulty in tuning a model that serves a global user base. He described the challenge of balancing helpfulness with neutrality as a “personality problem”.
By offloading tone decisions to the user, OpenAI attempts to sidestep the impossible task of aligning a single default personality with every user’s preference. This approach effectively shifts the responsibility for the “feel” of the conversation from the model’s fine-tuning team to the end-user.
Despite the new freedom in tone, the company has officially delayed its planned “Adult Mode” until 2026. Originally teased as a way to allow restricted NSFW content for age-verified users, the feature has faced setbacks.
Discussing the roadmap earlier this year, Altman suggested the platform could eventually “allow even more, like erotica for verified adults.”
Coinciding with heightened scrutiny following a wrongful death lawsuit involving a teenage user, the delay prompts a more cautious approach to NSFW content. OpenAI clarified that the new “Warmth” and “Enthusiasm” settings do not override core safety protocols or content filters.
Aligning with OpenAI’s broader strategy for its Atlas agentic browser, the push for more “human-like” interaction is strategic. As AI moves from passive chatbots to active assistants, the “personality” becomes a key user interface element.
Atlas weekly release is here, last one before the holidays:
– Multi-profile: Multiple browser profiles (‘personal’, ‘school’, ‘work’) for the same ChatGPT login; multiple ChatGPT logins coming soon
– Selectively import things like extensions or passwords, even after… pic.twitter.com/WzoBekYtws
– Adam Fry (@adamhfry) December 18, 2025
However, increasing autonomy brings security risks, particularly regarding how an “enthusiastic” agent might be manipulated. Security vulnerabilities such as prompt injection remain a critical concern for autonomous agents.
OpenAI CISO Dane Stuckey admitted in October that defending against these attacks is an ongoing battle. “Prompt injection remains a frontier, unsolved security problem, and our adversaries will spend significant time and resources to find ways to make ChatGPT agent fall for these attacks.”