Anomalous Agent Behavior Alerting With Tool-Call Baselines
How tool-call baselines catch AI agent attacks that standard monitoring misses.
Priya Subramaniam
Senior Editor & AI Security Correspondent
Priya Subramaniam
Priya Subramaniam spent eight years at a boutique cybersecurity consultancy auditing enterprise software supply chains before pivoting to full-time analysis of autonomous systems risk. She leads the publication's coverage of adversarial threats against AI agents and multi-agent orchestration pipelines.
7 stories
OTEL Tracing for AI Agent Session Forensics
OpenTelemetry spans reveal what AI agents decided to do, not what actually executed.
Priya Subramaniam
Tamper-Proof Audit Logging Architecture for AI Agent Actions
Agents need cryptographically verifiable logs to catch emergent harms across sessions.
Priya Subramaniam
OAuth 2.0 Limitations for Non-Human Agent Identities
OAuth's human-centered design breaks down when applied to autonomous AI agents.
Priya Subramaniam
MCP Supply Chain Campaign Anatomy
Authenticated access by design makes compromised MCP servers a novel supply chain threat.
Priya Subramaniam
AI Asset Sprawl as an Attack Surface Expansion Vector
AI agents multiply attack surface faster than security teams can inventory or control them.
Priya Subramaniam
Intent Drift Detection in Autonomous Agent Sessions
Researchers find autonomous agents gradually abandon their original goals, posing security risks.
Priya Subramaniam