Podcast: Will Agentic AI Bring Fantasia’s Sorcerer's Apprentice to Life?: A Conversation with Tracy Bannon

In this podcast episode, Michael Stiefel spoke with Tracy Bannon, a Senior Principal Software Architect and DevOps Advisor Senior at MITRE, about the role of artificial intelligence in software development and the associated risks. Bannon emphasized that AI is not a new threat to cybersecurity, but rather an evolution of existing threats.

She highlighted the challenges that come with testing agents that are non-deterministic and whose behavior changes as they are tested, as well as the cognitive load placed on software professionals due to the massive amount of code and documents AI agents can generate.

Bannon also discussed the danger of overreliance on artificial intelligence, leading to diminishing human skills, particularly in complex domains. She stated that the question of outsourcing intelligence to automated models is one that should be considered carefully. She emphasized that while individuals have control over their use of AI agents, the attack surface becomes immense when connecting these agents to personal and corporate systems.

She compared the situation to granting a person access to various keys, urging people to be cautious about the level of exposure they are willing to accept.

Moreover, Bannon shared an example of her own use of a browser extension for AI, where she authorized it to perform tasks while she stepped away. She warned that this action, although seemingly harmless, could open up a world of potential security risks. She encouraged listeners to think about their own use of AI agents and to take appropriate measures to safeguard their personal and professional information.