https://www.kitelogik.com/blog OPA/Rego policy enforcement at the tool execution layer. Deterministic. Auditable. Infrastructure-level. en Tue, 04 Aug 2026 00:00:00 GMT https://www.kitelogik.com/blog/agent-lifecycle-attacks https://www.kitelogik.com/blog/agent-lifecycle-attacks Four AI agent attack classes that live in the action layer — delegation, plans, budgets, and memory — and why input/output guardrails structurally miss them. Louis Bryson Tue, 04 Aug 2026 00:00:00 GMT Threat model Agent lifecycle OPA https://www.kitelogik.com/blog/jailbreak-under-policy-enforcement https://www.kitelogik.com/blog/jailbreak-under-policy-enforcement A jailbroken agent controls what it asks for, not what executes. Where that boundary actually falls — and the four things a successful jailbreak still buys an attacker. Louis Bryson Tue, 04 Aug 2026 00:00:00 GMT Prompt injection Threat model HITL https://www.kitelogik.com/blog/owasp-agentic-top-10-scored https://www.kitelogik.com/blog/owasp-agentic-top-10-scored Which of the ten 2026 agentic risks a deterministic policy layer actually enforces: three fully, six partially, and one it arguably makes worse. Louis Bryson Tue, 04 Aug 2026 00:00:00 GMT OWASP Threat model Policy as code https://www.kitelogik.com/blog/agent-incidents-no-attacker https://www.kitelogik.com/blog/agent-incidents-no-attacker The most expensive AI agent failures on record had no adversary — the agent did what it was told. What that means for a security model built on keeping attackers out. Louis Bryson Tue, 04 Aug 2026 00:00:00 GMT Incidents Agent safety Budgets https://www.kitelogik.com/blog/mcp-tool-poisoning-allowlist https://www.kitelogik.com/blog/mcp-tool-poisoning-allowlist A peer-reviewed benchmark found more capable models are more vulnerable to tool poisoning, not less. You cannot train your way out of this one. Louis Bryson Tue, 04 Aug 2026 00:00:00 GMT MCP Tool poisoning Supply chain https://www.kitelogik.com/blog/ai-act-delay-evidence-problem https://www.kitelogik.com/blog/ai-act-delay-evidence-problem High-risk obligations moved to December 2027. The event logging Article 12 requires still takes longer to retrofit than the extension buys you. Louis Bryson Tue, 04 Aug 2026 00:00:00 GMT EU AI Act Audit trail Record-keeping https://www.kitelogik.com/blog/your-first-agent-policy https://www.kitelogik.com/blog/your-first-agent-policy Write an agent policy in a constrained YAML DSL, compile it to Rego, unit-test it, and dry-run a single event — without learning Rego first. Louis Bryson Tue, 04 Aug 2026 00:00:00 GMT Rego Getting started Policy as code https://www.kitelogik.com/blog/where-to-put-the-gate https://www.kitelogik.com/blog/where-to-put-the-gate Three ways to wire policy enforcement into a Python agent — and the failure-mode difference that actually decides which one you want. Louis Bryson Tue, 04 Aug 2026 00:00:00 GMT Architecture Python Integration https://www.kitelogik.com/blog/budgets-are-a-security-control https://www.kitelogik.com/blog/budgets-are-a-security-control A ceiling that fires on magnitude halts a runaway agent without diagnosing it — and the counter feeding it is part of your trust boundary. Louis Bryson Tue, 04 Aug 2026 00:00:00 GMT Budgets Denial of wallet Agent safety https://www.kitelogik.com/blog/multi-agent-handoffs-trust-boundary https://www.kitelogik.com/blog/multi-agent-handoffs-trust-boundary Most frameworks treat an agent-to-agent handoff as a function call. It is a privilege transfer, and almost nothing checks it before it happens. Louis Bryson Tue, 04 Aug 2026 00:00:00 GMT Multi-agent Delegation OWASP