ChatGPT, which converses more frequently than with family, friends, or partners and shares sensitive personal information, performs requests and responds to questions all day long. Who are the people creating the personality, tone, and behavioral principles of such ChatGPT? Joanne Chang, head of OpenAI Labs, who researches how artificial intelligence (AI) interacts with humans, is at the core of this.

During a meeting on the 24th of last month at the OpenAI Korea office in Gangnam-gu, Seoul, Chang stated, "Eight hundred million users worldwide share not only their sensitive health and personal information but also their relationships with partners, family, and friends with ChatGPT and ask for solutions," adding, "I feel a great sense of responsibility as users rely on ChatGPT for almost everything in their lives."

Born in 1995, Chang is one of only two Korean women among the '100 Most Influential People in AI in 2025' announced by the American weekly magazine Time. She joined OpenAI in December 2021 after working at Dropbox and Google. She oversaw the behavior and policy of OpenAI's AI models and recently became the head of the core research organization, OpenAI Labs. This is her first interview with Korean media.



● "Designing Behavior for Coexistence of Humans and AI"

When setting the behavioral principles of AI, there is always a 'dilemma' between technology and ethics. She cited a theft incident as an example. Chang stated, "AI should refuse to answer requests for theft methods, but there is a dilemma when a store owner asks for theft patterns to respond to a theft incident," adding, "Ultimately, it's the same information asked in a different way, and we have to think about what response to give for the users." She emphasized, "When requests for help in making biochemical weapons come in, we refuse to answer for the sake of humanity as a whole," stressing that "how to design AI model behavior for the coexistence of humans and AI is the most important."

Establishing the personality of ChatGPT, which has now become a 'companion AI,' was not easy as it was an uncharted path. In August 2022, when Chang was in charge of planning the GPT-4 product, she was not satisfied with the model's tone and style. From then on, she assembled a team of talented individuals with excellent sensibilities, and over the past three years, they researched not just refining the tone but how the model should understand and respond to users to build trust. Through this, they created a new field called 'model behavior.' Chang stated, "On some days, we debate philosophical topics like 'to what extent should a user's religion be reflected in responses if they share it,' and on other days, we brainstorm with team members on how to respond when a user makes nonsensical remarks or how to provide comforting responses when a user says 'Hingㅠㅠ'."

● "Korea, a Country that Loves Technology, Will Become an AI Powerhouse"

Having lived in Korea until the second grade of elementary school, Chang reminisced, "Even after moving to the U.S., I watched all of Kim Yuna's figure skating performances," and expressed her delight, saying, "I'm glad OpenAI has opened a Korean branch to represent Korean users." She also mentioned, "During the AI model research process, I always paid attention to awkward or lacking expressions in Korean," and noted, "Korean users are very positive about new technologies and provide active feedback to OpenAI."

She also positively evaluated the Korean government's push for sovereign AI. Chang stated, "It will never be a structure where a specific foundation model monopolizes," adding, "If there is an AI model for Koreans that understands Koreans, it can lead to better development through competition." She further predicted, "It is too early to judge which specific model is the best or lagging behind," and "although a latecomer, Korea, a nation that loves technology, can catch up and become an AI powerhouse."

Regarding Sam Altman's leadership style, Chang stated, "Altman always tells employees, 'Set OpenAI's performance goals to ten million times the current level,'" implying that there are virtually no limits to his goals or ambitions. She also highlighted his friendly leadership as one of OpenAI's attractions. When Altman offered Chang a job, he personally gave her his phone number, saying, "Call anytime." Chang explained that even now, they frequently communicate whenever making important decisions, such as recruiting talent.

Popular News