AI Agent Governance & Safety Platforms cover the controls, observability, and policy tooling organizations need to deploy autonomous or assistant-style AI agents safely and compliantly. By 2026, enterprises are balancing rapid agentization across customer service, legal, finance, and productivity workflows with obligations around data protection, auditability, and systemic risk management. This topic compares how platforms implement those safeguards and where gaps remain. Key platform types include no-code agent builders (Cimba.AI, Duckie) that prioritize rapid, auditable creation of domain-specific agents; contact-center and conversational governance stacks (Observe.AI, Yellow.ai) that add real-time assist, QA automation, and voice agent controls; large-enterprise assistants and orchestration layers (IBM watsonx Assistant, Microsoft 365 Copilot) that integrate with enterprise data, role-based access, and multi-agent workflows; domain-specific compliance-focused solutions (Harvey for law/professional services); embedded CRM assistants (HubSpot Breeze); and specialized commerce/chatbots (Verifast). Gizzmo.ai was listed but its canonical site was unreachable during collection, so public details may be limited. Comparisons should focus on policy and safety features: provenance and data lineage, immutable audit logs, role-based access and approval workflows, model provenance and vendor risk controls, red‑teaming and test harnesses, runtime monitoring and drift detection, human‑in‑the‑loop escalation, and exportable compliance artifacts (risk assessments, model cards). Also relevant are integration boundaries (what enterprise systems agents can act on) and whether platforms support fine‑grained action controls. This synthesis aims to help procurement, security, and compliance teams evaluate tradeoffs between speed and control when selecting agent platforms, emphasizing measurable governance features over marketing claims.