Security
Audit trails that make AI agent work reviewable
May 15, 2026Security
Audit trails that make AI agent work reviewable
What an audit trail should capture so AI agent work stays inspectable after the task is done.
Facio Blog
Practical notes on human-reviewed AI agents.
Payload-powered product notes, security writing, HITL patterns, and operational guidance from the Facio runtime: long sessions, Placet approvals, audit trails, memory, providers, channels, tools, and Docker-first operations.
Showing 16-18 of 18 articles in Security.
Security
Credential boundaries for production AI agents
May 15, 2026Security
Credential boundaries for production AI agents
How to connect agents to useful systems without handing raw secrets to prompts, logs, or shell commands.
Security
250 Poisoned Documents: The AI Supply Chain Backdoor You Can't Patch
Draft date pendingSecurity
250 Poisoned Documents: The AI Supply Chain Backdoor You Can't Patch
250 poisoned documents — that's all it takes to create a permanent backdoor in any AI model, regardless of size or training data volume. Anthropic's 2025 research changed the threat model overnight. Here's what it means for enterprise AI security.