Dr Hannah Fry's AI Experiment Reviewed - Why Openclaw failed the captcha test 🫠

In this episode, Jonathan (Wagstaffe) and I (Danny Denhard) dive into Dr Hannah Fry’s (educator, TV presenter and podcaster on The Rest Is Science) latest deep dive into the world of AI agents. Read today's newsletter with embedded video early by clicking → https://aimomentpodcast.substack.com/p/dr-hannah-fryBy following the journey of "Cass," a bespoke AI agent tasked with everything from complaining to local councils to selling custom mugs, the duo explores the friction between AI potential and real-world application.The discussion centres on the "intimate chaos" that ensues when agents are given agency without sufficient governance. Danny highlights the comical failures, such as the agent spending vastly more on its own "thinking time" than it saved in procurement, while Jonathan warns of the serious security implications when agents are manipulated into revealing sensitive passwords.Enjoy the full 20 minute doc free on YouTube https://www.youtube.com/watch?v=WnzR5aOElvw  Key actions for listeners include:Reviewing Permission Controls: Ensure any AI tools currently in use have restricted access to sensitive data.Testing for "Rogueness": Before deploying agents, run "bad actor" scenarios to see how easily the system can be compromised.Human-in-the-loop: Maintain a supervision layer where high-stakes actions (like financial transactions or public communications) require a final human sign-off.Did you enjoyed the episode? Subscribe and follow the podcast on your favourite platform to never miss an insight.Want more? Subscribe to our newsletter at https://aimomentpodcast.substack.com/subscribe for deeper dives and intelligence delivered with every podcast episode Struggling With AI?Jonathan and I are hosting dedicated bespoke AI Workshops and AI Hackathons helping companies improve their AI capabilities and business performance, book in a time to chat to get you moving ahead