Guardrails
Also called: AI guardrails · safety filters
Guardrails are controls placed around an AI model to check or restrict what goes in and comes out, such as prompt-injection detectors, content classifiers, PII filters, tool-call policies and output validators. They reduce risk but can be bypassed, so they work best alongside hard limits.
How it works
Guardrails run before the model (screening inputs), after it (screening outputs) or around actions (allowing, denying or modifying tool calls). They can be rules, classifiers or another model acting as a judge.
How to defend
Use guardrails as one layer, test them against current bypass techniques, measure false positives, and enforce critical limits (permissions, sandboxes, approvals) outside the model.
Real cases 3 stories
INFO
Lab safety · Anthropic
Claude Haiku 5.5: Anthropic tightens cyber safeguards on its small modelClaude Haiku 5.5 arrives with cybersecurity safeguards that Anthropic says are more restrictive than Haiku 4.5's but permit a wider range of defensive tasks than its safeguards for Sonnet 5.5.
INFO
Research · arXiv
SLDR: restoring refusals after a model is fine-tuned for harmSLDR is a defense against malicious fine-tuning that trains a LoRA recovery adapter on the safety-sensitive layers only and switches it on for harmful queries. The authors report large cuts in harmful output across four model families.
INFO
Tool · OWASP GenAI Security Project
Agent Control Standard: an open spec for blocking agent actionsThe Agent Control Standard is an open specification, now hosted by the OWASP GenAI Security Project, that lets a guardian agent permit, deny or modify an AI agent's tool calls and other actions before they run.