Do Chatbots Walk the Talk of Responsible AI?

Public commitments don't always translate to practice: The study found "significant gaps between corporate rhetoric and practice" among leading chatbot companies, meaning their stated responsible AI principles on websites and in documentation often don't match what the chatbots actually do when tested.

Multiple evaluation methods reveal inconsistencies: By examining company websites, technical documentation, and directly testing the chatbots themselves, researchers uncovered discrepancies that wouldn't be visible from looking at just one source—suggesting companies may communicate responsible practices they haven't fully implemented.

Problem spans major industry players: The analysis covered four major chatbots (ChatGPT, Gemini, DeepSeek, and Grok), indicating this isn't an isolated issue with one company but a broader pattern across the industry's leading systems.

Transparency gap matters for accountability: The disconnect between what companies say about responsible AI and what their systems actually do undermines public trust and makes it harder for users, regulators, and researchers to hold companies accountable for their stated commitments.