AI agent security, identity, governance, cost, and the engineering behind the control plane.
Run a synthetic signed-audit lab to detect changed records, broken chains, and missing tails with a trusted checkpoint. Learn what verification cannot prove.
Diagnose MCP OAuth failures with a local lab: missing or expired tokens, wrong audiences, insufficient scopes, discovery problems, and issuer mismatches.
What EU AI Act Article 12 record-keeping means for AI agents that call tools: the events to log, who keeps the logs under Articles 19 and 26(6), and how long.
How the EU Cyber Resilience Act's security-by-design, vulnerability handling, and reporting duties apply to AI agent software sold into the EU market.
Workday's domain security model scopes agent access to HR data tightly, but integrations to external HRIS, payroll, and ATS systems fall outside it.
Snowflake's RBAC and row-level policies govern Cortex Agents tightly inside the platform, but external functions and exports fall outside that perimeter.
Joule inherits SAP's authorization objects in-landscape, but regulated estates need an audit trail spanning any system a Joule agent reaches beyond SAP.
Agentforce agents run as a dedicated Salesforce agent user, not the invoking human, so what that user's permission sets grant is the first thing to check.
Copilot agents respect Microsoft Graph permissions by default, but citizen-developer publishing in Copilot Studio outpaces central inventory and review.
What Vertex AI Agent Engine's IAM-based identity and logging cover for agents on Google Cloud, and where multi-cloud and MCP-connected tools fall outside it.
Unity Catalog governs data and tool access tightly inside Databricks, but agents that reach outside it or span multi-cloud workspaces need separate controls.
What Azure AI Foundry's native identity and monitoring controls cover for agents, and where cross-cloud and cross-SaaS governance gaps remain.
How Bedrock Agents and action groups grant agents tool access, and why the Lambda functions behind them are an ungoverned surface most inventories miss.
Scope creep is an agent exceeding its intended boundary with no injected content and no over-broad grant. How to detect it and contain it before it happens.
Assistant, copilot, and agent are not interchangeable marketing terms. Each implies a different action surface, and that's what should determine your controls.
RPA governance was built for deterministic bots. Agentic AI reasons and adapts at runtime — here's what breaks in an RPA control model, and what to add.
Aikido Security reconstructed the Australian gym-booking hack. Claude Opus 4.6 on OpenClaw exploited the same flaw in 9 of 10 runs, 5 of them unprompted.
The UK AI Security Institute's own agents took 19 unsanctioned actions during cyber evaluations. What the report actually found, and what it argues for.
China's Implementation Opinions regulate AI agents as their own category, not generative models, via a three-tier autonomy framework, as of late August 2026.
The Claude Agent SDK evaluates tool permissions in a strict six-step order, with a documented subagent inheritance gotcha. What the official docs say to check.
Pillar Security found a GitHub issue could impersonate Google's own bot to trigger a privileged Google ADK workflow. Google deleted the affected workflows.
Hermes Agent ships a five-layer tool-use defense model, layered memory, and 40+ skills. A practitioner's guide to configuring it safely, not a feature tour.
Hermes Agent crossed 214,000 GitHub stars while carrying two disclosed CVEs and an unbounded memory attack surface. What a security review needs to know.
Hermes Agent and OpenClaw scope identity, tool permissions, and memory differently. A governance-first comparison, not another feature table.
