Recent studies have identified key gaps in the delegation mechanisms that allow agentic AI to interact with websites on behalf of users. These limitations pose challenges for the effective implementation of AI in critical task scenarios.
The primary issue stems from the restricted access control protocols currently employed by many websites. This lack of robust mechanisms complicates the AI's ability to operate efficiently and securely.
The ongoing research aims to enhance user trust in agentic AI systems while improving their operational efficiency. Addressing these access control challenges is essential for the broader adoption of AI technologies in critical applications.