Resources
Research, whitepapers, and reporting on AI security and the agentic era.
The State of AI Data Path Security
Moving from posture scans to runtime execution governance
November 2026
AI agents now act on regulated enterprise data, but the audit trail of what they do is incomplete. This paper explains why governance must move to the data layer and act at the moment of execution, where the request, query and result meet and harm can still be prevented. It sets out four requirements: where, when, context and proof. Learn how signed action records bring risk management and zero trust to the data layer, support audit and regulatory evidence, and inform six RFP questions you can apply to any vendor.
Your assistant crosses platforms. Your governance does not.
August 2026
For security leaders approving assistants that reach more than one enterprise system, and platform engineers making that approval deployable. The paper shows how cross-platform assistants create exposure no individual SaaS vendor can close, and how to determine whether your estate can produce the record an investigation needs.
Read the WhitepaperDeploying Agentic AI: Assessing what your agents actually did
August 2026
For security and compliance leaders approving AI agents against production data, and the architects who decide how those agents connect. The paper maps the four ways agentic AI reaches data, shows what each topology makes observable, and helps teams determine which records their architecture can produce.
Read the WhitepaperAI Agents Are Rewriting Compliance
The Evidence Layer Your Auditor Will Ask For
July 2026
Shows how new AI statutes and existing audit frameworks converge on the same missing record: what AI agents access in data. The dated paper discusses how signed, per-query execution evidence maps onto relevant frameworks and examines Claude Access Attestation.
Read the WhitepaperGoverning Claude in the Enterprise: The Missing Data Access Layer
July 2026
Claude is embedded across the enterprise, and Anthropic's Compliance API gives security vendors programmatic access to conversation records and policy hooks. This paper maps the visibility gap between an agent platform's record and the data-path record, then explains how the two can be reconciled.
Read the WhitepaperIntent Computing: The Security Paradigm for the Agentic Era
January 2026
Defines behavioral baselining for enterprise security by comparing requests with an agent's established pattern of behavior.
Read the Whitepaper