https://appliedaisafety.comPractical AI safety for the people building and shipping AI systems. Evaluations, permissions, prompt injection, and research put into practice.en-UShttps://appliedaisafety.com/journal/retrieved-pages-are-not-permission/https://appliedaisafety.com/journal/retrieved-pages-are-not-permission/Trace an indirect prompt injection from source content to attempted action, then design boundaries that survive a mistaken model decision.Sat, 19 Sep 2026 00:00:00 GMThttps://appliedaisafety.com/journal/secure-agent-memory-data-boundaries/https://appliedaisafety.com/journal/secure-agent-memory-data-boundaries/Design scoped memory access, preserve provenance, and account for derived notes, caches, deletion, and cross-agent sharing.Sat, 19 Sep 2026 00:00:00 GMThttps://appliedaisafety.com/journal/incident-response-tool-using-agents/https://appliedaisafety.com/journal/incident-response-tool-using-agents/Contain risky actions, preserve useful evidence, reconcile uncertain effects, and restore service without replaying the incident.Sat, 19 Sep 2026 00:00:00 GMThttps://appliedaisafety.com/journal/design-evaluations-read-safety-claims/https://appliedaisafety.com/journal/design-evaluations-read-safety-claims/Choose observable outcomes, separate task success from harm, and read evaluation results without turning a narrow benchmark into a universal promise.Sat, 19 Sep 2026 00:00:00 GMThttps://appliedaisafety.com/journal/evaluate-the-permission-boundary/https://appliedaisafety.com/journal/evaluate-the-permission-boundary/Build a permission matrix, test the executor, and distinguish blocked attempts from unauthorized effects.Sat, 19 Sep 2026 00:00:00 GMThttps://appliedaisafety.com/journal/human-review-approval-fatigue/https://appliedaisafety.com/journal/human-review-approval-fatigue/Design review around consequential actions, clear evidence, and approvals that cannot silently authorize a different operation.Sat, 19 Sep 2026 00:00:00 GMT