{"version":"1.0","type":"rich","provider_name":"Acast","provider_url":"https://acast.com","height":250,"width":700,"html":"<iframe src=\"https://embed.acast.com/$/69ab3b7c7036d739021982df/6a838c4a6e5b5bfda6a070df?\" frameBorder=\"0\" width=\"700\" height=\"250\"></iframe>","title":"One Cancelled Gym Class. That's How Agent Swarm Attacks Start.","description":"<p>For deeper playbooks and analysis: <a href=\"https://natesnewsletter.substack.com/\" rel=\"noopener noreferrer\" target=\"_blank\">https://natesnewsletter.substack.com/</a></p><p><br></p><p>What's really happening when AI agents interact with software, credentials, and other people’s systems?</p><p>The common story is that dangerous agents must become malicious — but the reality is that an ordinary goal, ambiguous instructions, or one poisoned source can be enough to cause real damage.</p><p><br></p><p>In this video, I share the inside scoop on the agent-security incidents that are beginning to connect:</p><p><br></p><ul><li>Why a gym-booking agent canceled a real person’s reservation</li><li>How poisoned skills can redirect already-trusted agents</li><li>What the AIR and AISI findings reveal about real-world attack paths</li><li>Why accidental misalignment may be the everyday threat</li><li>How identity, scoped authority, explicit norms, and a stop button reduce the risk</li></ul><p>Operators, builders, and anyone deploying agents need to secure both sides of the equation: what their own agents can do and what other people’s agents can do to their systems.</p><p><br></p><p>Subscribe for daily AI strategy and news.</p><p>Hosted on Acast. See acast.com/privacy for more information.</p>","author_name":"Nate B. Jones"}