Posture tells you what agents can reach.
Agent DLP governs what they actually do.
ArgusAI already maps every agent and the data it can touch. Agent DLP enforces similar policies at runtime, so what you intend and what actually happens stay in step.
01
Sit inline at your gateway
Deploy the Bedrock Data interceptor into your own AWS account and attach it to an existing AgentCore gateway. Your agents and tools stay as they are.
Native hooks for AgentCore and LiteLLM
Runs in your account
Under 10ms per call
02
Inspect request and response
A request carrying data a tool should not receive is blocked before it arrives. A response carrying data the agent should not see is withheld before it reaches the model.
Both directions of every call
Data, identity, and policy together
Uses your existing Data Types
03
Set policy, then trust the record
Allow some data types, redact others, or observe and record for later review. Every call lands in a continuous, audit ready log.
Target, action, data types, verdict
EU AI Act, ISO/IEC 42001, state rules
Configured with your other policies
A complete data profile for every agent
Bedrock Data pairs the data an agent can reach with what it actually does: defense in depth for agents. Posture narrows the blast radius; runtime enforcement closes the remaining gap.
Live without re-architecting anything
Setup is two steps, and neither your agents nor your tools have to change.
01
Deploy the interceptor
Drop the Bedrock Data interceptor Lambda into your own AWS account. It stays in your environment and reads only what it needs.
02
Attach it to your gateway
Set it as the request and response interceptor on your AgentCore gateway. Traffic starts flowing through inspection right away.
03
Shape policy over time
Start in observe mode to learn what agents touch, then tighten to redact or block by data type when you are ready.
Traditional DLP was built for people. Agent DLP was built for agents.
Legacy DLP watches employees moving files and never sees an agent's tool calls or the speed they run at. Agent DLP works at the layer where agent data actually moves.
Control at the moment of action
From knowing what an agent can access to governing what it does. Every operation is allowed, redacted, or blocked in real time, judged on the data, the identity, and the policy together.
Runs inside the gateway you have
Bedrock Data plugs into your existing agent infrastructure through native hooks. There is no separate proxy to stand up, route around, or maintain, and your source of truth on data stays intact.
Every decision in full context
Bedrock Data reads the data an agent touches, the identity behind it, and the policy that applies, all at once, using classifications your team already trusts.
Audit ready by default
Regulators now expect proof of what your AI systems did, not just what you intended. Agent DLP produces the continuous record the EU AI Act, ISO/IEC 42001, and state AI rules ask for.
With Bedrock Data, we can let agents work on the data that makes them valuable, because we control what they do with it the moment they act.
- Blocked. A support agent's get_customer call returning an SSN and card number is stopped on the spot.
- Redacted. A response with email and phone number comes back with those fields removed, and the agent keeps working with the rest.
- Observed. Lower-risk activity is recorded for later review, with no interruption to the agent.