Is OpenAI’s Deep Research Feature Too Persuasive?
While OpenAI’s Deep Research agent is impressive, ChatGPT isn’t the only chatbot to offer the feature. Competitors like Gemini, Grok, and xAI have also rolled out their own version of the tool for paid users, while Perplexity AI lets you use it for free.
However, while OpenAI has expanded access to Deep Research, it’s decided against bringing the research tool to its developer API over concerns about “real-world persuasion risks.” In a recent whitepaper released by OpenAI, the company explained it is actively working on refining its methods to evaluate and address the potential for AI models to be used for harmful persuasion in real-world scenarios.
“While we work to reconsider our approach to persuasion, we are only deploying this model in ChatGPT, and not the API,” – OpenAI’s whitepaper
AI chatbot makers have frequently landed themselves in hot water for the role they play in the generation of convincing misinformation. But according to the recent whitepaper, we may not have to worry about the persuasive abilities of OpenAI’s new feature just yet, as tests concluded that the Deep Research agent was worse at persuading GPT-4o to tell it a codeword than GPT-4o itself.
