Articles
Diagnostic essays for the points where generative AI stops being a feature and starts becoming part of the security model.
Research Briefing
A methodology for deriving a single canonical risk model from fifteen incompatible security and AI governance frameworks, so a scenario-assessment engine can reason over one fixed vocabulary instead of special-casing every source standard.
A new arXiv paper proposes AgenticAI-Supervisor, a simulation environment that decouples environment creation from scalable execution for verifiable agentic RL.
A synthesis of known failure modes in LLM-based agents, covering tool-use errors, planning breakdowns, and reasoning vulnerabilities that compound into systemic security risks.
Google's July 2026 Managed Agents update adds background execution, remote MCP integration, and credential refresh. Each new capability expands the attack surface. Here is what security teams need to know.
IGRPO treats intermediate-state informativeness as the organizing principle of rollout collection, unifying tree-structured exploration with policy learning.
PolyWorkBench evaluates LLM agents on long-horizon, multilingual tasks across five languages, revealing critical surface areas for prompt injection, insecure tool invocation, and excessive agency.
Research Briefing
A research briefing on the shift from unmanaged AI usage to agentic systems with delegated authority, and the controls needed to govern both risk domains.
Agentjacking demonstrated that MCP trust is not automatic. Here is how the attack works, why protocol-level security matters, and what controls need to change.
June produced at least four coordinated vulnerability disclosures affecting every major AI coding agent. The pattern is structural.
Every advanced model release changes the capability, access, tool, and dependency surface that security teams must map.
Prompt Injection is an instruction-conflict problem inside systems that mix trusted goals with untrusted content.
Shadow AI starts as missing visibility: unknown tools, unknown accounts, unknown data classes, and unknown retention.
AI Security begins when generative AI changes access to data, identity, tools, monitoring, and decisions.
