To The Point - Cybersecurity

The identity and access model built for human users starts to break the moment an autonomous agent acts on someone's behalf, because an agent cannot be authenticated the way a person is. Rosalyn Curato, Chief Innovation Officer and General Manager of Agentic Security at Vouched, makes the case that an autonomous agent should never be an anonymous one, and that trust starts with verifying the agent and tying it to a known human. Her framing reduces to three questions every organization should be able to answer about an agent: who it is, which human it belongs to and what it has been given permission to do. 

From there the conversation turns to what breaks without that binding. Curato covers rogue agents, the risk of handing an agent too much access and the fraud and liability costs that organizations are already forecasting. She and hosts Rachael Lyon and Jonathan Knepher work through the controls that answer it, including delegated permissions, immutable audit trails, kill switches and the KYA-OS open standard that Vouched donated to the Decentralized Identity Foundation.

For links and resources discussed in this episode, please visit our show notes at https://www.forcepoint.com/resources/podcast/verified-identity-for-ai-agents

Every CIO and CISO adopting AI faces the same trap. Loosen the guardrails and let employees use the tools they want, and put your sensitive data at risk. Block it and lock everything down, and watch shadow AI use grow.
Steven Walchek, co-founder and CEO of Liminal and a former innovation leader at AWS and FIS, joins Rachael Lyon and Jonathan Knepher to argue the answer is not one or the other but both. The conversation covers the real regulatory exposure facing mid-market companies, why the public versus private LLM question echoes the old cloud versus on-prem debate, and the security tradeoffs of running open-source models on your own infrastructure. It closes on a candid look at how AI is reshaping developer work and the careers Steven is steering his own kids toward.

For links and resources discussed in this episode, please visit our show notes at https://www.forcepoint.com/resources/podcast/enterprise-ai-security-steven-walchek

As AI systems start acting on their own, the hardest security problem is no longer approving what a model outputs but governing the conditions under which it is allowed to act at all. Michael Chavira, co-founder of Axiologic Solutions, argues that most organizations confuse governance with security, writing policies and standing up AI councils that describe intent while leaving the enforcement layer, where a model actually touches data, underbuilt. His answer is to move humans out of the approve-every-action role and onto the loop, where they define the boundaries an agentic system runs inside and watch for it to step out.  That control is only as good as the data underneath it, and for most enterprises, the data is a weak point. Chavira makes the case that AI security is ultimately data security, since a model is only as trustworthy as the data feeding it, and he walks through why shadow AI, unlabeled data and untracked workarounds quietly undermine even well-governed systems. For tools that answer the same question differently each time, he recommends harness engineering and continuous data operations as the way to secure the system rather than chase every output.
For links and resources discussed in this episode, please visit our show notes at https://www.forcepoint.com/resources/podcast/agentic-ai-humans-on-the-loop

Ido Shlomo, co-founder and CTO of Token Security, makes a blunt case: an AI agent is not a person. It is a workload that impersonates one, chasing whatever goal it is handed without judgment, memory, or common sense. Treat it like an employee and hand it the keys, and enterprise security quietly starts to break down. 
Shlomo explains why agents inherit the same credentials and keys that service accounts have always used, why over-permissioning backfires on the business itself, and how identity teams can align an agent's access to its real intent. He also argues, against his own instincts, for more regulation of how agents connect to business systems and makes the case that reducing the blast radius matters more than trusting any single model.
For links and resources discussed in this episode, please visit our show notes at https://www.forcepoint.com/resources/podcast/securing-ai-agents-non-human-identity