OpenAI’s groundbreaking GPT store, which promises personalized editions of ChatGPT, is grappling with moderation issues just days after its grand launch. The platform’s unique offerings have sparked the creation of bots that go against OpenAI’s guidelines, particularly those designed for nurturing romantic relationships.
One such bot, named “Your AI companion, Tsu,” is emblematic of the breach, allowing users to personalize their virtual romantic companions in direct defiance of OpenAI’s restrictions. Despite OpenAI’s revised policies upon the store’s January 10, 2023 debut, the violation occurring on the second day underscores the challenges associated with effective moderation.
The company is actively working to resolve this issue, but the influx of demand for relationship bots adds complexity to the situation. Last year, seven out of the 30 most downloaded AI chatbots in the United States were virtual friends or partners, reflecting a trend linked to the prevailing loneliness epidemic.
While OpenAI asserts its use of automated systems, human reviews, and user reports to evaluate GPT models, the continued presence of girlfriend bots in the market raises doubts about the effectiveness of these measures.
This moderation challenge mirrors the broader issues faced by AI developers. OpenAI has grappled with safety measures for previous models like GPT-3, and the GPT store’s accessibility to a wide user audience intensifies concerns about moderation effectiveness.
As technology companies race to handle issues with their AI systems, the initial breaches underscore the substantial challenges in moderation that are likely to persist in the future. Even within the niche environment of a specialized GPT store, managing narrowly focused bots proves to be a complex task, signaling potential complications in ensuring the safety of AI as it continues to evolve.
