Explore the significant security risks posed by Agentic AI and quantum computing, from data breaches of sensitive information to the breakdown of modern cryptography. Learn proactive strategies for governing AI and preparing for a quantum-safe future.
Jason Martin of Permiso Security discusses the exponential rise of AI agents in enterprises and the urgent security challenges they present. He covers the concept of Non-Human Identity (NHI), applying Zero Trust principles to ephemeral and over-permissioned agents, and outlines key attack vectors like prompt injection and data poisoning, while also exploring the potential of defensive AI to enhance security operations.
Laura Deming, co-founder and CEO of Until, discusses the ambitious goal of making reversible cryopreservation a reality. She outlines the company's vision, starting with preserving human organs to revolutionize transplantation and progressing towards whole-body medical hibernation, allowing patients to "pause" biological time and await future cures. The conversation delves into the core scientific challenges, reframing cryopreservation as an engineering problem of heat transfer and managing ice formation, and explores why this promising field has been historically underfunded.
Amey Shukla from the University of Connecticut presents a novel system for biometric key derivation that closes the long-standing gap between the theory and practice of device-level authentication. The talk introduces a practical fuzzy extractor system, "Zeta then Lock," which, combined with an integrated machine learning feature extractor, achieves 105 bits of entropy with a 92% true accept rate for iris biometrics, overcoming the "more errors than entropy" problem that plagued previous designs.
Uma Ramadoss from AWS explains the core concepts of Agentic AI, differentiating it from standard AI workflows. The session covers how to build agentic systems on AWS using services like Bedrock and Step Functions, and open-source frameworks like Strands SDK, emphasizing practical architecture, context enrichment, and the importance of verification.
A summary of the Google DeepMind team's discussion on AlphaGenome, their unified DNA sequence-to-function model. It covers the scientific motivation, the engineering breakthroughs in processing long DNA sequences at high resolution, the addition of complex biological modalities like splicing and contact maps, and the future direction of the research.