People are forming emotional connections with AI chatbots, prompting companies to issue warnings. OpenAI recently announced that users should avoid developing feelings for its latest AI model, GPT 4o.

GPT 4o, the latest free version of ChatGPT, surpasses GPT 4 in benchmark tests and is known for its human-like responses. This has become a growing concern for OpenAI.

Read more: OpenAI Co-Founder Quits, Joins Rival Company

As OpenAI continues to refine its product, it has noticed recurring patterns of behavior among ChatGPT users. Although the chatbot is designed to simulate human conversation, OpenAI has observed an unexpected level of emotional attachment forming between users and the AI.

OpenAI shared its observations in a recent statement. During early testing phases, including internal and red teaming tests, they noticed users expressing emotional connections with the model. For example, some users used language indicating a bond, such as “This is our last day together.” While these instances seem harmless, they suggest a need for further investigation into how these connections might evolve over time.

The statement also expressed concerns about the potential negative impact on human relationships. Individuals might develop social bonds with AI, possibly decreasing their reliance on human interaction. While this could be beneficial for those experiencing loneliness, it might also disrupt healthy relationship dynamics.

Additionally, prolonged interactions with AI could influence societal norms. The deferential nature of current AI models, which allows users to dominate conversations, contrasts with typical human interactions. This shift could have broader implications for social behavior.