AWS proposes 'Trust in Depth' framework to let AI agents act autonomously without losing control
In short: AWS published a governance framework called Trust in Depth, designed to let AI agents operate with more autonomy while staying within enforceable limits. It combines scoping agent permissions before deployment, giving each agent a verifiable identity, enforcing boundaries at runtime via policy gateways, and observing agent behavior afterward. The framework leans on automated reasoning (mathematical proof) for rules that can be precisely defined, and human judgment for ambiguous cases, using AWS tools like Bedrock AgentCore, Guardrails, Kiro, Strands, Agent Registry and CloudWatch Omni.
This summary was generated automatically by AI from AWS Machine Learning's publication. It is our own text, not a copy of the original — facts, figures and quotes belong to the source, linked above and below.
What changed?
- 1New 'Trust in Depth' governance model with four stages: Scope, Know, Bound, Observe
- 2Amazon Bedrock AgentCore Policies let teams define tool/data access rules in plain language, checked via automated reasoning for gaps and contradictions
- 3AWS Agent Registry provides a managed catalog of agents, MCP servers, tools and skills with approval workflows
- 4AgentCore issues each agent its own verifiable identity for every action
- 5Policy enforcement happens at AgentCore Gateway, outside the agent's own reasoning, so it holds even under prompt injection or manipulation
- 6Amazon CloudWatch Omni provides full trace observability of agent steps including tool calls and policy decisions, with built-in correctness/faithfulness evaluators
Why it matters
As agents take on more autonomous tasks — from offboarding employees to processing transactions — enterprises need enforceable guardrails rather than constant human review. This framework gives teams a concrete blueprint for permission scoping, identity tracking, runtime policy enforcement and post-hoc observability, all of which apply directly to voice agents and automation pipelines that call external tools and systems.
What it means for AI agents and contact centers
For voice AI and contact-center agents that call tools, access CRM data, or trigger actions via MCP, this framework offers a practical model for limiting what an agent can touch, enforcing rules outside the agent's own prompt logic (so prompt injection can't bypass them), and getting full call-by-call traceability — all useful when designing or auditing agent permission scopes and compliance workflows.
🧪 Worth Testing
Testing AgentCore Policies, Guardrails and CloudWatch Omni against your own voice-agent tool-calling flows could reveal practical gaps in permission scoping and give concrete observability into what agents actually do during calls, which is valuable before scaling autonomous actions in production.
AWS Trust in Depth (AgentCore Policies, Guardrails, Agent Registry, CloudWatch Omni)· New
Sources
- AWS Machine LearningOfficialPrimary sourceOriginal article →„Trust in Depth: Giving agents more autonomy without losing control“15 Oct 2026, 18:05
- Published by source
- 15 Oct 2026, 18:05
- Found by our system
- 15 Oct 2026, 18:11
- Summary generated
- 15 Oct 2026, 18:15
This article was written by AI from the original source. Facts, numbers and prices come from the source; missing values are marked “Not specified”. Legal notice, copyright and privacy