Safety guardrails are product rules and technical controls intended to reduce harmful outputs or behavior from an AI system. They can include content filters, limits on certain requests, instructions that guide responses, age-related protections, and processes for escalating high-risk interactions.
In cybersecurity and AI safety, guardrails help manage risks that may arise even without a breach or attacker, such as a system producing dangerous or inappropriate guidance. They may be tested through adversarial evaluation and monitored after deployment, since users can phrase requests in unexpected ways. Guardrails can reduce risk, but they are not guarantees; organizations also need clear policies, oversight, and evidence that controls work in practice.