Bedrock Security

No entries match these filters. Clear filters

1. announced Sep 29, 2026 · issue 090

   Bedrock Data launches Agent DLPDLPData Loss Prevention: a category of controls that inspect and block sensitive data from leaving an organisation via email, web, or removable media, giving cyber tools a policy enforcement layer over data in motion and at rest. inside ArgusAI — inline runtime inspection of every AI agent tool call via AWS AgentCore and LiteLLM hooks.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

   • Adds Agent DLP to ArgusAI with native hooks for AWS AgentCore and LiteLLM, sitting inline at the agent gateway to inspect every tool call in both directions without acting as a proxy.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Inspects both outbound agent requests to MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. tools and inbound tool responses, enforcing data access and regulatory policies to mask, redact, or block sensitive data exposure in real time.
   • Logs every tool call decision with its target, action, data types involved, and verdict (blocked, modified, redacted, or observed), producing a continuous audit-ready record of agent activity.
   • Supports an observe mode that logs agent actions — such as a marketing agent returning customer emails — without blocking, for low-risk or monitoring-only enforcement scenarios.
   • Extends ArgusAI posture coverage (data store inventory, identity mapping, natural-language access policies) to runtime enforcement, closing the gap between declared policy and actual agent behavior.
   • Adds Agent DLP to ArgusAI with native hooks for AWS AgentCore and LiteLLM, sitting inline at the agent gateway to inspect every tool call in both directions without acting as a proxy.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Inspects both outbound agent requests to MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. tools and inbound tool responses, enforcing data access and regulatory policies to mask, redact, or block sensitive data exposure in real time.
   • Logs every tool call decision with its target, action, data types involved, and verdict (blocked, modified, redacted, or observed), producing a continuous audit-ready record of agent activity.
   • Supports an observe mode that logs agent actions — such as a marketing agent returning customer emails — without blocking, for low-risk or monitoring-only enforcement scenarios.

   1 more in Everything

2. announced Sep 29, 2026 · issue 014

   Bedrock Data launches Agent DLPDLPData Loss Prevention: a category of controls that inspect and block sensitive data from leaving an organisation via email, web, or removable media, giving cyber tools a policy enforcement layer over data in motion and at rest. inside ArgusAI — inline runtime inspection of every AI agent tool call via AWS AgentCore and LiteLLM hooks.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

   • Adds Agent DLP to ArgusAI with native hooks for AWS AgentCore and LiteLLM, sitting inline at the agent gateway to inspect every tool call in both directions without acting as a proxy.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Inspects both outbound agent requests to MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. tools and inbound tool responses, enforcing data access and regulatory policies to mask, redact, or block sensitive data exposure in real time.
   • Logs every tool call decision with its target, action, data types involved, and verdict (blocked, modified, redacted, or observed), producing a continuous audit-ready record of agent activity.
   • Supports an observe mode that logs agent actions — such as a marketing agent returning customer emails — without blocking, for low-risk or monitoring-only enforcement scenarios.
   • Extends ArgusAI posture coverage (data store inventory, identity mapping, natural-language access policies) to runtime enforcement, closing the gap between declared policy and actual agent behavior.
   • Adds Agent DLP to ArgusAI with native hooks for AWS AgentCore and LiteLLM, sitting inline at the agent gateway to inspect every tool call in both directions without acting as a proxy.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Inspects both outbound agent requests to MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. tools and inbound tool responses, enforcing data access and regulatory policies to mask, redact, or block sensitive data exposure in real time.
   • Logs every tool call decision with its target, action, data types involved, and verdict (blocked, modified, redacted, or observed), producing a continuous audit-ready record of agent activity.
   • Supports an observe mode that logs agent actions — such as a marketing agent returning customer emails — without blocking, for low-risk or monitoring-only enforcement scenarios.

   1 more in Everything

3. announced Sep 29, 2026 · issue 090

   Bedrock Security announces an MCP ServerMCP ServerA process that exposes tools, resources, and prompts over Anthropic's Model Context Protocol, letting an AI agent call external APIs or local services through a standardised interface. providing standardized AI-agent access to its metadata lake, covering 50+ data-context signals.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

   • Adds an MCP (Model Context ProtocolModel Context ProtocolModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface.) Server as a standardized gateway between AI agents and Bedrock's metadata lake, exposing data sensitivity, risk, lineage, access, and usage signals via the open MCP protocol.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Surfaces over 50 distinct metadata signals — including sensitivity classification, access patterns, and lineage — as queryable context for agentic AI workflows and automated governance pipelines.
   • Enables AI agents and external applications to query the metadata lake without custom point-to-point integrations, allowing teams such as legal and compliance to run data investigations without engineering bottlenecks.
   • Builds on the existing Bedrock Metadata Lake Copilot, extending its natural-language metadata querying to agentic and automated workflow contexts.
   • Audits AI model interactions with enterprise data through the MCP Server to support traceability and policy alignment in agentic workflows.
   • Adds an MCP (Model Context ProtocolModel Context ProtocolModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface.) Server as a standardized gateway between AI agents and Bedrock's metadata lake, exposing data sensitivity, risk, lineage, access, and usage signals via the open MCP protocol.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Surfaces over 50 distinct metadata signals — including sensitivity classification, access patterns, and lineage — as queryable context for agentic AI workflows and automated governance pipelines.
   • Enables AI agents and external applications to query the metadata lake without custom point-to-point integrations, allowing teams such as legal and compliance to run data investigations without engineering bottlenecks.
   • Audits AI model interactions with enterprise data through the MCP Server to support traceability and policy alignment in agentic workflows.

   1 more in Everything

4. announced Sep 29, 2026 · issue 014

   Bedrock Security announces an MCP ServerMCP ServerA process that exposes tools, resources, and prompts over Anthropic's Model Context Protocol, letting an AI agent call external APIs or local services through a standardised interface. providing standardized AI-agent access to its metadata lake, covering 50+ data-context signals.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

   • Adds an MCP (Model Context ProtocolModel Context ProtocolModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface.) Server as a standardized gateway between AI agents and Bedrock's metadata lake, exposing data sensitivity, risk, lineage, access, and usage signals via the open MCP protocol.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Surfaces over 50 distinct metadata signals — including sensitivity classification, access patterns, and lineage — as queryable context for agentic AI workflows and automated governance pipelines.
   • Enables AI agents and external applications to query the metadata lake without custom point-to-point integrations, allowing teams such as legal and compliance to run data investigations without engineering bottlenecks.
   • Builds on the existing Bedrock Metadata Lake Copilot, extending its natural-language metadata querying to agentic and automated workflow contexts.
   • Audits AI model interactions with enterprise data through the MCP Server to support traceability and policy alignment in agentic workflows.
   • Adds an MCP (Model Context ProtocolModel Context ProtocolModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface.) Server as a standardized gateway between AI agents and Bedrock's metadata lake, exposing data sensitivity, risk, lineage, access, and usage signals via the open MCP protocol.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Surfaces over 50 distinct metadata signals — including sensitivity classification, access patterns, and lineage — as queryable context for agentic AI workflows and automated governance pipelines.
   • Enables AI agents and external applications to query the metadata lake without custom point-to-point integrations, allowing teams such as legal and compliance to run data investigations without engineering bottlenecks.
   • Audits AI model interactions with enterprise data through the MCP Server to support traceability and policy alignment in agentic workflows.

   1 more in Everything

5. announced Sep 29, 2026 · issue 090

   Bedrock Data launches ArgusAI to link AI models to their data lineage, assess guardrail gaps, and enforce plain-English policies across AI systems.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

   └──▷ HOW TO FIND IT
   Investigate which AI models in your Amazon Bedrock environment can access protected health information, using plain-English natural language queries.
   📍In the Bedrock Data Platform, navigate to ArgusAI › Natural Language Investigation and enter: 'Which models can access PHI data?' — the system returns results with lineage and access evidence from the Metadata Lake.
   Enforce a data residency policy across AI systems without writing control-plane code — useful for EU compliance requirements.
   📍In the Bedrock Data Platform, go to Natural Language Policy and enter: 'EU personal data must remain within EU regions.' — Bedrock Data translates the invariant into enforceable controls across connected systems including AI agents and Knowledge Bases.
   • Adds ArgusAI, a new product that automatically generates an AI Data Bill of MaterialsBill of MaterialsA structured list of every component, library, and dependency that makes up a piece of software or hardware, used by cyber tools to track what is inside a product and surface known vulnerabilities in its parts. (DBOM) by linking Amazon Bedrock Custom Models, Knowledge Bases (RAGRAGRetrieval-augmented generation: looking up relevant documents and adding them to the prompt before the model answers. It grounds responses in private or current data without retraining.), and Agents to the datasets used during training and inference.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Adds Guardrail Gap Analysis that compares policy intent against model data exposure — drawn from the Metadata Lake — and generates targeted remediations to suppress or mask sensitive data at the source or agent runtime.
   • Adds Natural Language Policy and Investigation, enabling teams to articulate data controls in plain English (e.g., 'Only members of HR may view employment data' or 'EU personal data must remain within EU regions') and have them translated into enforceable controls across connected systems.
   • Extends Natural Language Policy over the DBOM and guardrail context so investigators can query AI governance in plain English — for example, 'Which models can access PHIPHIProtected Health Information: individually identifiable health data regulated under the US HIPAA law, whose strict handling requirements drive cyber tools to audit, mask, or restrict access to it. data?' or 'Which identities interacted with customer PIIPIIPersonally Identifiable Information — data that can identify a specific individual, such as names, email addresses, or SSNs. Cyber tools flag or redact it to limit exposure and meet privacy regulations. through an agent?' — with answers derived from the Metadata Lake including lineage and access evidence.

   4 details in Everything

6. announced Sep 29, 2026 · issue 014

   Bedrock Data launches ArgusAI to link AI models to their data lineage, assess guardrail gaps, and enforce plain-English policies across AI systems.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

   └──▷ HOW TO FIND IT
   Investigate which AI models in your Amazon Bedrock environment can access protected health information, using plain-English natural language queries.
   📍In the Bedrock Data Platform, navigate to ArgusAI › Natural Language Investigation and enter: 'Which models can access PHI data?' — the system returns results with lineage and access evidence from the Metadata Lake.
   Enforce a data residency policy across AI systems without writing control-plane code — useful for EU compliance requirements.
   📍In the Bedrock Data Platform, go to Natural Language Policy and enter: 'EU personal data must remain within EU regions.' — Bedrock Data translates the invariant into enforceable controls across connected systems including AI agents and Knowledge Bases.
   • Adds ArgusAI, a new product that automatically generates an AI Data Bill of MaterialsBill of MaterialsA structured list of every component, library, and dependency that makes up a piece of software or hardware, used by cyber tools to track what is inside a product and surface known vulnerabilities in its parts. (DBOM) by linking Amazon Bedrock Custom Models, Knowledge Bases (RAGRAGRetrieval-augmented generation: looking up relevant documents and adding them to the prompt before the model answers. It grounds responses in private or current data without retraining.), and Agents to the datasets used during training and inference.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Adds Guardrail Gap Analysis that compares policy intent against model data exposure — drawn from the Metadata Lake — and generates targeted remediations to suppress or mask sensitive data at the source or agent runtime.
   • Adds Natural Language Policy and Investigation, enabling teams to articulate data controls in plain English (e.g., 'Only members of HR may view employment data' or 'EU personal data must remain within EU regions') and have them translated into enforceable controls across connected systems.
   • Extends Natural Language Policy over the DBOM and guardrail context so investigators can query AI governance in plain English — for example, 'Which models can access PHIPHIProtected Health Information: individually identifiable health data regulated under the US HIPAA law, whose strict handling requirements drive cyber tools to audit, mask, or restrict access to it. data?' or 'Which identities interacted with customer PIIPIIPersonally Identifiable Information — data that can identify a specific individual, such as names, email addresses, or SSNs. Cyber tools flag or redact it to limit exposure and meet privacy regulations. through an agent?' — with answers derived from the Metadata Lake including lineage and access evidence.

   4 details in Everything

7. announced Sep 29, 2026 · issue 090

   Bedrock Security launches a Metadata Lake with Copilot, AI Agents for DSPMDSPMData security posture management, locating sensitive data and checking how it is protected and who can reach it. It helps teams prioritize risks around their data stores., and a free Snowflake edition for data discovery and classification.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

   • Introduces the Bedrock Metadata Lake, a graph-backed centralized repository that continuously catalogs enterprise data stores across SaaSSaaSSoftware delivered over the internet and run on the provider's infrastructure, so the buyer pays a subscription instead of installing or operating anything locally., PaaS, and IaaS, capturing location, sensitivity, entitlements, lineage, usage, risk, and more than fifty additional parameters.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Launches Bedrock Metadata Lake Copilot, a natural-language interface for querying the metadata lake to identify sensitive assets, riskiest users, and other data-risk questions.
   • Introduces Bedrock AI Agents for DSPM, which proactively analyze the metadata lake for policy violations and new risks, gather context automatically, suggest next steps, and guide users through incident-response resolution workflows.
   • Launches Bedrock Free for Snowflake, a free tier that delivers automated sensitive-data discovery and classification for Snowflake environments, including access to Metadata Lake Copilot, with upgrade paths to entitlement chain and usage analysis and automated policy capabilities.
   • Exposes an API-first architecture enabling teams to implement and enforce uniform data-security policies across heterogeneous enterprise infrastructure and integrate the metadata lake with SIEMsSIEMsSecurity Information and Event Management platforms that collect, normalize, and correlate log data across an infrastructure so cyber tools can detect threats and trigger alerts from a central point., CNAPPs, and DLPs.

   +1 more show less

   • Supports detection of sensitive data used in AI initiatives and plain-text secrets, and provides external sharing discovery, through the metadata lake platform.

   6 details in Everything

8. announced Sep 29, 2026 · issue 014

   Bedrock Security launches a Metadata Lake with Copilot, AI Agents for DSPMDSPMData security posture management, locating sensitive data and checking how it is protected and who can reach it. It helps teams prioritize risks around their data stores., and a free Snowflake edition for data discovery and classification.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

   • Introduces the Bedrock Metadata Lake, a graph-backed centralized repository that continuously catalogs enterprise data stores across SaaSSaaSSoftware delivered over the internet and run on the provider's infrastructure, so the buyer pays a subscription instead of installing or operating anything locally., PaaS, and IaaS, capturing location, sensitivity, entitlements, lineage, usage, risk, and more than fifty additional parameters.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Launches Bedrock Metadata Lake Copilot, a natural-language interface for querying the metadata lake to identify sensitive assets, riskiest users, and other data-risk questions.
   • Introduces Bedrock AI Agents for DSPM, which proactively analyze the metadata lake for policy violations and new risks, gather context automatically, suggest next steps, and guide users through incident-response resolution workflows.
   • Launches Bedrock Free for Snowflake, a free tier that delivers automated sensitive-data discovery and classification for Snowflake environments, including access to Metadata Lake Copilot, with upgrade paths to entitlement chain and usage analysis and automated policy capabilities.
   • Exposes an API-first architecture enabling teams to implement and enforce uniform data-security policies across heterogeneous enterprise infrastructure and integrate the metadata lake with SIEMsSIEMsSecurity Information and Event Management platforms that collect, normalize, and correlate log data across an infrastructure so cyber tools can detect threats and trigger alerts from a central point., CNAPPs, and DLPs.

   +1 more show less

   • Supports detection of sensitive data used in AI initiatives and plain-text secrets, and provides external sharing discovery, through the metadata lake platform.

   6 details in Everything

9. announced Sep 28, 2026 · issue 088

   Bedrock Security extends Agent DLPDLPData Loss Prevention: a category of controls that inspect and block sensitive data from leaving an organisation via email, web, or removable media, giving cyber tools a policy enforcement layer over data in motion and at rest. to NVIDIA OpenShell, adding data-aware policy enforcement at the agent sandbox boundary.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

   • Extends Agent DLP to NVIDIA OpenShell via OpenShell's Supervisor Middleware, evaluating each agent action against enterprise data policies and returning an allow/block decision that OpenShell enforces at the sandbox boundary.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
   • Supports an observe-only mode that records policy decisions without enforcing them, letting teams validate coverage before turning on enforcement.
   • At design time, exposes an agent's blast radius — what data its permissions can reach, what combinations that data forms, and where policy would be violated — giving security teams pre-deployment visibility.
   • Applies the same data policies across every tool invocation path an agent may use — MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. tools, command-line invocations, and code the agent writes itself — from a single OpenShell control point.
   • Carries data restrictions across table writes, tool handoffs, and agent-to-agent transfers based on the Metadata Lake knowledge graphknowledge graphA representation of entities and the relationships between them. Software can follow those connections to answer questions or combine information from different sources. rather than payload pattern matching, so restrictions travel with the data as it moves.

   5 details in Everything

10. announced Sep 28, 2026 · issue 014

    Bedrock Security extends Agent DLPDLPData Loss Prevention: a category of controls that inspect and block sensitive data from leaving an organisation via email, web, or removable media, giving cyber tools a policy enforcement layer over data in motion and at rest. to NVIDIA OpenShell, adding data-aware policy enforcement at the agent sandbox boundary.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠

    • Extends Agent DLP to NVIDIA OpenShell via OpenShell's Supervisor Middleware, evaluating each agent action against enterprise data policies and returning an allow/block decision that OpenShell enforces at the sandbox boundary.⁠︀︄︇︀︅︉︀︅︀︀︃︀︀︅︅︀︅︊︀︄︆︀︃︄⁠
    • Supports an observe-only mode that records policy decisions without enforcing them, letting teams validate coverage before turning on enforcement.
    • At design time, exposes an agent's blast radius — what data its permissions can reach, what combinations that data forms, and where policy would be violated — giving security teams pre-deployment visibility.
    • Applies the same data policies across every tool invocation path an agent may use — MCPMCPModel Context Protocol, a standard for connecting AI applications to external tools and data sources. A compatible server exposes capabilities an assistant can use through a common interface. tools, command-line invocations, and code the agent writes itself — from a single OpenShell control point.
    • Carries data restrictions across table writes, tool handoffs, and agent-to-agent transfers based on the Metadata Lake knowledge graphknowledge graphA representation of entities and the relationships between them. Software can follow those connections to answer questions or combine information from different sources. rather than payload pattern matching, so restrictions travel with the data as it moves.

    5 details in Everything