Agent Safety

Artificial intelligence can act on her own, but she needs limits. These lessons show how we set boundaries and keep her aligned with our goals.

• Guardrails and oversight – Constraining and monitoring agent behavior.
• Human-in-the-loop agents – Agents guided by human review and approval.