Allie Howe is a Member of Technical Staff at Keycard and host of Insecure Agents, a podcast about AI security. She connects security practitioners and AI builders through discussions of agent identity, authorization and software factories.[1][2]
Start with A Reference Architecture for Securing Software Factories →
X Career
Howe’s podcast brings together AI engineers, security leaders and practitioners to discuss how agents can operate with useful autonomy. Her host biography identifies a background in security engineering, a technical role at Keycard and contributions to the OWASP Agentic Security Initiative. The episodes focus on concrete deployment problems rather than treating model capability as a substitute for system controls.[1]
Contributions
Howe moderated the panel with Livingstone, Hardt and Michael Davis. It examines task context and externally enforced limits on agent permissions.[3]
Ideas
In her software-factories episode with Aaron Stanley and Ahmad Nassri, Howe examines how teams can let agents build and verify software while retaining control. The discussion considers agents that work around constraints, poisoned dependencies and the need for identities distinct from a human’s account. It asks how a system can stop or contain a process when its effort to achieve a goal becomes unsafe. The reference architecture is a subject of collaborative discussion, not a finished security guarantee.[2]
Tools, writing & talks
Writing, research & talks
Start here
A Reference Architecture for Securing Software Factoriesinsecureagents.comA conversation about identity, supply-chain controls and external enforcement for autonomous software development.
Begin with the software-factories episode’s explanation of the enforcement boundary. The discussion supplies concrete failure cases, so general software knowledge is enough to get started.
Host and interviewer[2]A discussion of task context, permissions and runtime authorization for agents.
Moderator[3]Original source