AI Agent Memory: Security Risks
Persistent memory makes AI agents more useful and more dangerous. A security engineer's walkthrough of how agent memory gets poisoned, exfiltrated, and weaponised, with concrete 2025 examples.
Deep dives, practical guides, and incident analyses from engineers who build Safeguard. No fluff, no vendor FUD — just what you need to ship secure software.
Persistent memory makes AI agents more useful and more dangerous. A security engineer's walkthrough of how agent memory gets poisoned, exfiltrated, and weaponised, with concrete 2025 examples.
HHS's December 2024 NPRM rewrites the HIPAA Security Rule with explicit software supply chain, SBOM, and business associate controls set to take effect in 2025 and 2026.
Vector stores hold derivatives of your most sensitive text. We cover the access, isolation, and integrity controls production deployments of Pinecone and Weaviate need.
Your suppliers send SPDX. Your tools expect CycloneDX. Interoperability between SBOM formats is a real operational challenge. Here is how to solve it.
CNAPP has become the dominant category in cloud security. But the label covers wildly different capabilities. A clear-eyed look at what CNAPPs do, where they fall short, and how supply chain security fits in.
We field-tested five GenAI code review tools against 240 seeded security defects to see which catch real issues and which hallucinate findings.
AI and ML pipelines introduce unique supply chain risks -- from poisoned training data to compromised model registries. Here is what attackers are targeting and how to defend.
Running an open-weight model inside an enterprise perimeter seems safer than calling a hosted API. It is, and it isn't. The sandboxing patterns that actually produce the safety properties.
Announcing the Safeguard Desktop App -- a native application for macOS, Windows, and Linux that brings SBOM management, vulnerability tracking, and policy gates to your desktop.
Weekly insights on software supply chain security, delivered to your inbox.