All news with #amazon bedrock tag

🚀 Amazon Bedrock AgentCore introduces a managed harness (preview), an AgentCore CLI, and prebuilt skills to help developers build and iterate on agents faster. The managed harness runs the complete agent loop in a per-session microVM with filesystem and shell access, supports model-agnostic switching, and lets teams export orchestration as Strands-based code. Filesystem persistence (preview) enables suspending and resuming sessions, and the CLI integrates with AWS CDK to deploy agents with governance and audibility.

🔎 AWS Deadline Cloud now includes an AI-powered troubleshooting assistant that analyzes failed render jobs to diagnose root causes and recommend fixes. The assistant examines logs and metrics for issues like missing assets, software errors, configuration mismatches, and resource constraints, drawing on a pre-trained knowledge base covering Deadline Cloud and popular DCC apps. It runs inside your AWS account via Amazon Bedrock and is available in all regions that support Deadline Cloud.

🚀 Claude Opus 4.7 is now available in Amazon Bedrock, delivering Anthropic’s most capable Opus release with improvements across coding, professional knowledge work, visual understanding, and long-running task handling. Served via Bedrock’s next-generation inference engine, Opus 4.7 offers enterprise features such as zero operator data access, dynamic traffic routing, and improved scalability. The model enhances agentic coding, systems engineering, long-horizon reasoning, and high-resolution image support, and is available in select AWS Regions.

🔍 Amazon Bedrock now supports cost allocation by IAM principal in AWS Cost and Usage Report 2.0 (CUR 2.0) and Cost Explorer. Customers can tag IAM users and roles with attributes like team, project, or cost center, activate those tags as cost allocation tags, and either include caller identity in a CUR 2.0 export or filter by tags in Cost Explorer. This capability is available in all AWS commercial Regions where Amazon Bedrock is offered.

🔐 Unit 42 found the AgentCore starter toolkit auto-creates overly permissive IAM roles that grant wildcard access to Bedrock AgentCore and ECR resources. The default deployment enables an “Agent God Mode” scenario where a compromised agent can exfiltrate container images, retrieve other agents’ MemoryIDs, invoke code interpreters, and read or poison memories across an entire AWS account. AWS updated documentation to warn these roles are intended for development; Unit 42 recommends creating scoped, least-privilege roles and auditing ECR, memory, and invoke permissions.

🖱️ Amazon announced that Bedrock AgentCore Browser now supports OS-level interaction capabilities to automate browser workflows beyond the Chrome DevTools Protocol (CDP). The update adds mouse and keyboard operations, full desktop screenshots, and handling for native system dialogs and print flows. It targets AI agent developers, test automation engineers, and teams building LLM-powered web interaction tools. The feature is enabled by default across all AgentCore Browser instances in the 14 supported AWS Regions.

🛡️ At AWS, decades of scaled security operations combine with new AI collaborations to proactively harden critical systems. Through Project Glasswing and Anthropic’s Claude Mythos Preview, AWS runs continuous AI-driven code reviews and provides gated research previews via Amazon Bedrock. Complementary offerings include AWS Security Agent for autonomous penetration testing and Bedrock guardrails and Automated Reasoning to enforce enterprise controls and reduce risk.

🔒 Amazon Bedrock now offers Claude Mythos Preview in a gated research preview as part of Project Glasswing. Anthropic's most advanced model to date demonstrates state-of-the-art capabilities across cybersecurity, software coding, and complex reasoning, identifying sophisticated vulnerabilities and showing exploitability in large codebases with less manual guidance. Access is limited to an allow-list in US East (N. Virginia) through Bedrock; AWS account teams will contact approved organizations.

🤖 AWS Cost Explorer now integrates Amazon Q generative AI so you can query your AWS cost and usage data using natural language. Suggested prompts and a new 'Ask Question' button let users pose common or custom questions while Cost Explorer automatically updates charts, filters, and groupings to match the query. When Amazon Q uses additional datasets like pricing or anomaly detection, visual outputs appear in an artifacts panel. The conversation remains contextual for follow-up analysis without switching tools.

🔒 This Unit 42 analysis evaluates Amazon Bedrock Agents' multi-agent collaboration from a red-team perspective. The researchers demonstrate a chain of reconnaissance and exploitation—detecting operating mode, enumerating collaborator agents, delivering attacker-controlled payloads, and triggering tool actions—when Bedrock Guardrails and pre-processing are disabled. The report confirms no vulnerabilities in Bedrock itself and emphasizes mitigations such as Bedrock Guardrails, input validation, scoped agent capabilities, and the principle of least privilege.

🎙️ Amazon Bedrock Data Automation now supports custom vocabulary through the new Data Automation Library, enabling organizations to supply domain-specific word lists to improve speech recognition and transcription accuracy for audio and video. The capability supports display-form mapping (for example, rendering “electrocardiogram” as “ECG”) and covers 11 languages. It targets sectors such as healthcare, legal, financial services, media, and contact center analytics and is provided at no additional charge.

🔒 Amazon Bedrock Guardrails now supports centralized, organization-wide enforcement through cross-account safeguards, enabling security teams to apply configurable safety controls from a single management account. AWS reports these safeguards can block up to 88% of harmful multimodal content and help filter hallucinated model outputs, removing the need to configure guardrails per account. The capability is available in all supported commercial and GovCloud regions and can be managed via the AWS Console or APIs.

🔒 SageMaker Data Agent now supports cross-region inference profiles for Japan (JP-CRIS) and Australia (AU-CRIS) via Amazon Bedrock. Inference requests originating in Asia Pacific (Tokyo) and Asia Pacific (Sydney) are processed entirely within their respective geographies, helping customers meet data residency and sovereignty requirements. Data Agent continues to provide conversational data exploration, Python and SQL code generation, troubleshooting, and analytics inside SageMaker Unified Studio Notebooks and the Query Editor, with traffic routed exclusively over the AWS Global Network.

🔒 Amazon Bedrock now supports structured outputs in AWS GovCloud (US) Regions. Structured outputs enables foundation models to produce consistent, schema-compliant, machine-readable responses tailored for government and regulated workloads. This reduces the need for custom validation logic, lowers operational overhead, and minimizes failed requests and retries. The capability is generally available in all commercial and GovCloud regions where Amazon Bedrock is supported.

🎯 Amazon Bedrock AgentCore Evaluations is now generally available to deliver automated, continuous and on-demand quality assessment for AI agents. The feature provides online evaluation to sample and score live production traces and on-demand evaluation for programmatic tests in CI/CD pipelines and interactive workflows. It includes 13 built-in evaluators covering response quality, safety, task completion and tool usage, plus Ground Truth and customizable LLM- or code-based evaluators.

🔒 This post explains how financial institutions should augment traditional security frameworks with AI-specific controls when deploying agentic AI. It emphasizes two foundational capabilities—comprehensive observability of agent workflows and fine-grained tool access controls—to preserve explainability and accountability. The author presents seven design principles and actionable implementation guidance, referencing SR 11-7 and practical AWS tooling such as Amazon Bedrock AgentCore and monitoring integrations.

🤖 AWS Step Functions now supports 28 additional AWS service integrations and over 1,100 new API actions, enabling direct orchestration of a broader set of services without writing custom integration code. New integrations include Amazon Bedrock AgentCore for invoking AI agent runtimes and Amazon S3 Vectors for document ingestion workflows. The update also adds support for AWS Lambda durable execution APIs to enable idempotent durable function invocations and management of durable executions from workflows. These enhancements are generally available in all Regions where Step Functions is offered.

🖼️ Amazon Bedrock now offers Palmyra Vision 7B from Writer, a multimodal 7B-parameter model that interprets and generates text from images via the Bedrock API. Trained on PixMo, a dataset of one million high-quality image-text pairs, the model supports visual question answering, handwritten text extraction, chart interpretation, and image-based comprehension for enterprise workflows. It is available in select AWS Regions to simplify deployment of multimodal applications.

🔒 Amazon now enables Bedrock AgentCore to apply Chrome Enterprise policies to AgentCore Browser and to accept custom root Certificate Authority (CA) certificates for both AgentCore Browser and Code Interpreter. Administrators can leverage 100+ configurable browser policies — such as URL restrictions, disabling password managers, download controls, and kiosk-mode restrictions — to enforce compliance for AI agents. Custom root CA support permits secure TLS connections to internal services and corporate proxies that use enterprise-signed certificates, helping agents operate within strict security environments.

📦 Amazon Bedrock AgentCore Runtime now offers managed session storage in public preview. When configured, each session receives a persistent directory at your specified mount path so agents can read and write files as usual while the runtime transparently replicates data to durable storage. On stop, data is flushed during graceful shutdown and, when you resume with the same session ID, a new microVM mounts the same storage so source files, installed packages, build artifacts, and git history are preserved. The feature supports standard Linux filesystem operations, provides up to 1 GB per session, retains data for 14 days of idle time, and confines storage communication to a single session for isolation.