All news with #bedrock agents tag

🛡️ Security advisories from Amazon Web Services, Google, and Vercel detail vulnerabilities in agent harnesses that allowed caller-supplied or forged tool-call data to reach execution without a verifying model turn. AWS patched its managed Bedrock AgentCore service, Google fixed ADK for Python in v2.5.0, and Vercel released fixes for its Codex and OpenCode harness packages. Each vendor's path differed in attack surface and required conditions, and mitigations focus on validating and binding tool invocations to authenticated model events.

🛡️ Amazon Bedrock AgentCore introduces temporal policies for stateful agent authorization and rate limiting to control AI traffic. Temporal policies evaluate requests in the context of prior actions within a session, enabling workflow sequencing, exact argument matching, human approvals, and data freshness checks. Rate limiting provides per-user and per-group controls via OAuth or AWS IAM to cap requests, token usage, and concurrent connections for improved downstream availability. Documentation, a blog announcement, and the Dogwood reference implementation offer regional details and guidance.

🔎 Amazon Bedrock AgentCore now sends traces, prompts, structured logs, and standard output to a single per-agent Amazon CloudWatch log group, consolidating previously separated telemetry. This removes the need to search across multiple log groups and enables scoping of IAM policies and CMK encryption at the agent level. New agents created from July 20, 2026 use this by default in supported regions; existing agents can enable it via an environment variable and ADOT 0.17.1+.

🚀 Amazon Bedrock AgentCore has increased default runtime quota limits to enable larger-scale agent workloads. The update raises active concurrent sessions to 5,000 in US East (N. Virginia) and US West (Oregon), and 2,500 in other supported Regions. All Regions now support 200 agent interactions per second and 25 new sessions per second, improving out-of-the-box throughput for agent deployments. Customers should review the AgentCore Quotas documentation and Developer Guide for details.

🚀 Amazon Bedrock AgentCore is now available in four additional AWS Regions: Asia Pacific (Bangkok), Asia Pacific (Malaysia), Europe (Milan), and Europe (Spain). AgentCore is a platform to build, connect, and optimize agents, enabling engineers to ship agents quickly with any framework and model while enforcing security at the infrastructure layer. The launch brings runtime, identity and access control, policy management, session persistence, tool connectivity, and observability to these regions, reducing latency for local end users.

🛡️ AWS Security Hub CSPM introduces the AI Security Best Practices standard, offering 31 automated controls to detect misaligned AI resources. The standard evaluates Amazon Bedrock, Bedrock AgentCore, and Amazon SageMaker workloads against recommended configurations without manual rule creation. It covers domains like network isolation, encryption, VPC placement, KMS usage, private registries, and authorization, producing findings to help teams remediate issues. Available in all Regions where Security Hub CSPM operates, including GovCloud (US) and China.

🔒 AWS announces general availability of AWS WAF protection for Amazon Bedrock AgentCore Gateway, enabling protection of agentic AI workloads from common web exploits and abuse. You can associate an AWS WAF protection pack with your AgentCore Gateway to enforce IP-based access controls, rate-based throttling, and AWS Managed Rule Groups including Bot Control. Configure protections once at the Gateway and have them applied consistently to all targets behind it.

🔒 Amazon Bedrock AgentCore Memory now supports cross-account access, enabling multi-account architectures where memory resources and consuming agents span AWS accounts. Administrators can attach resource-based policies to memory resources to grant principals in other accounts permission to call memory data plane APIs by referencing the full memory ARN. Cross-account delivery destinations let memory resources stream payloads and events to Amazon S3, Amazon SNS, and Amazon Kinesis Data Streams in separate accounts. This capability is available in all Regions where AgentCore Memory is supported.

🔒 Check Point and AWS are collaborating to secure enterprise AI agents by integrating Amazon Bedrock AgentCore with Check Point AI Security. This partnership extends AgentCore’s identity, gateway, registry, and policy capabilities with runtime behavioral protections that monitor agent actions across models, tools, data, and applications. The integration enables policy-driven governance, visibility into agent deployments, and prevention of misuse such as prompt injection or unintended data exposure.

🚀 Amazon Bedrock AgentCore announces general availability of its managed agent harness, enabling teams to deploy production-grade agents in minutes. The harness handles orchestration, tool execution, session isolation, persistent memory, failure recovery, and context management so customers define agents via configuration rather than coding the loop. It supports any model, mid-session model switching, integrated security and observability, and exports to code for custom orchestration, and is available today in all AWS Commercial Regions where AgentCore is offered.

📌 Amazon Bedrock AgentCore Memory now allows applications to attach metadata values directly to short-term memory events so those values are preserved exactly when converted into long-term memory records. Previously, metadata had to be inferred by the LLM during extraction. Setting a metadata key to STRICTLY_CONSISTENT ensures the provided value is unchanged, prevents merging across different values, and lets you group events for department scoping, compliance boundaries, and multi-tenant isolation. The feature supports up to three strictly consistent STRING keys per strategy and is available in all AWS Regions where AgentCore Memory is supported.

🔎 Amazon Bedrock AgentCore now offers Web Search generally available in US East (N. Virginia). The fully managed tool provides agents with current web knowledge, combining an Amazon-operated web index and structured knowledge graph data to return high-value excerpts, entity data, and verified facts. Web Search integrates with AgentCore via the Model Context Protocol (MCP) and keeps data residency within your secured AWS environment with zero data egress.

🖥️ Amazon Bedrock AgentCore Runtime introduces the InvokeAgentRuntimeCommandShell API, providing a persistent, PTY-backed terminal over WebSocket into running agent sessions. This complements existing one-shot execution via InvokeAgentRuntimeCommand and delivers a full terminal experience inside an isolated microVM with features like colors, tab completion, Ctrl+C, resize, and automatic reconnect. Developers hosting coding agents (for example, Claude Code, OpenAI Codex, Amazon Kiro) can now authenticate, drop into the agent microVM, inspect files, run ad-hoc commands, and debug while retaining session state across reconnects. Each interactive session uses a runtime session ID and shell ID for resume; up to 10 concurrent shells are supported per runtime.

🤖 AWS Step Functions now integrates with the Amazon Bedrock AgentCore managed harness (preview) to add AI agent reasoning steps to workflows. The integration lets you declare agents via configuration, run agents in parallel or sequence, add human approvals, and view execution history with agent inputs, outputs, token usage, and CloudWatch links. You can reuse or create harnesses from Workflow Studio, apply per-invocation overrides, and persist agent context with session IDs. The harness preview and integration are available in select regions and standard Step Functions and Bedrock pricing applies.

🧭 Amazon SageMaker AI introduces multi-turn reinforcement learning (RL), a serverless model customization method for fine-tuning models on multi-step, agentic tasks. The feature trains models against users' agent environments, rewarding entire decision sequences to improve task accuracy of smaller, cost‑effective models versus larger general-purpose models. It integrates with Amazon Bedrock AgentCore Runtime and other deployment targets, and handles rollout orchestration, trajectory collection, training, and checkpoints, with MLflow tracking and evaluation metrics. Multi-turn RL runs serverlessly and is available in SageMaker Studio and the SageMaker Python SDK, supporting several foundation models in specific regions.

🔒 This post shows how SaaS providers can use Amazon Bedrock AgentCore resource-based policies to control multi-tenant access to a shared AgentCore Runtime and Runtime endpoint. It walks through two tenant scenarios: cross-account access for Example Corp and VPC-restricted access for AnyCompany, demonstrating how to apply resource-level Allow and explicit Deny conditions. The article covers required IAM permissions, example policy files, and verification steps to ensure network- and identity-based constraints are enforced.

🔐 Amazon Bedrock AgentCore Identity now lets customers reference existing AWS Secrets Manager secret ARNs directly in Credential Providers. Previously, secrets were service-managed and created by AgentCore Identity, limiting tagging, CMK encryption, and governance controls. Customers can now create and manage secrets with their own policies and then reference the ARN without changing runtime behavior. This feature is GA in 14 AWS Regions.

🔒 Amazon explains how AgentCore Gateway enforces a centralized authorization layer between autonomous agents and external tools, treating the LLM as an untrusted actor. Policies are expressed in the open-source Cedar language for readability, bounded execution, and mathematical analyzability, enabling deterministic enforcement and formal verification during policy authoring and attachment. A neuro-symbolic workflow translates natural-language rules into Cedar, validates them with Cedar Analysis, and enforces decisions at runtime to constrain tool invocations and filter unavailable actions.

🔒 The updated AWS User Guide to Governance, Risk, and Compliance for Responsible AI Adoption provides Financial Services customers practical GRC guidance for deploying AI responsibly. It covers governance, risk management, compliance, data and model management, and AI agent oversight, and maps these considerations to AWS capabilities. The guide highlights services such as Amazon Bedrock AgentCore, Bedrock Guardrails, Bedrock Agents, SageMaker Autopilot, and SageMaker Model Monitor. It complements existing AWS responsible AI and Well-Architected resources and is available on the AWS Whitepaper portal.

💳 Amazon Bedrock AgentCore now offers a preview of AgentCore payments, enabling AI agents to autonomously discover and pay for APIs, MCP servers, web content, and other agents. Built with Coinbase and Stripe, the feature manages wallet authentication, x402 protocol negotiation, stablecoin payment execution, and proof delivery without interrupting an agent's reasoning loop. Developers can attach a Coinbase CDP or Stripe Privy wallet, set session-level spending limits enforced at the infrastructure layer, and observe every transaction through AgentCore's existing logs, metrics, and traces. The Coinbase x402 Bazaar MCP server is accessible via AgentCore Gateway, and the preview is available in four AWS Regions.