Arctic Wolf Labs reported that attackers hosted their own instances of the open-source AI penetration-testing platform AntiFox on Alibaba Cloud and used them to analyze compromised systems and hunt for secrets. In observed incidents, attackers planted ASP.NET web shells on exposed Windows IIS servers, used valid IAM credentials and EC2 roles to explore AWS environments, and accessed, copied, and deleted three S3 objects. AWS said its services were not compromised and that the activity involved credentials stolen from customer applications, and it advised rotating exposed credentials, limiting permissions, requiring IMDSv2, and enabling Amazon GuardDuty.
AWS Machine Learning Blog describes three checks for a PyTorch DistributedDataParallel training loop that detect silent data corruption before corrupted values can spread through gradients and replica weights. The checks cover non-finite gradients, ring replay of micro-batches, and post-update weight fingerprint comparison; when triggered, the workflow maps a suspect rank to its node and asks SageMaker HyperPod to replace it. AWS says it tested the approach on a four-instance cluster with 32 NVIDIA H100 GPUs using software-injected faults.
AWS CEO Matt Garman said space-based AI data centers are unlikely to be economically feasible or practical within 10 to 20 years. He cited launch cost and frequency, payload limits, heat rejection, radiation, and maintenance as barriers, while acknowledging that the problems are technically solvable. Elon Musk continues to advocate SpaceX’s space-compute plans and has claimed that purpose-built aluminum radiators can dissipate 1,300 watts per square meter.
AWS released aws-emulator, an official tool for testing against local clones of multiple AWS services without provisioning a paid API key. Simon Willison reported that the Rust implementation was written by a large group of agents and recorded 29,684 commits in its first three weeks. The emulator currently requires Docker and runs on macOS or Windows.
AWS outlined an event-driven architecture for synchronizing AWS Agent Registry records with MCP server tools and schemas. AWS CloudTrail captures UpdateAgentRuntime API calls, Amazon EventBridge routes matching events, and AWS Lambda queries the MCP server, compares its tools with the registry, and updates the matching record. The design supports single-account deployments and cross-account event forwarding. Updated records return to DRAFT status and must pass the approval workflow before becoming discoverable again. The sample requires an existing matching registry record, does not automatically register previously unknown servers, has no manual force-sync path, and does not retry when a cold MCP server returns zero tools.
AWS released AWS Emulator, a free Rust-based application that provides local API endpoints for more than 200 AWS services, allowing developers to test applications without an AWS account or connection to live services. AWS said agents using Kiro wrote the product code, while automated fidelity tests and merge gates check its behavior against real AWS services. The source is available on GitHub under primarily CC0 licensing. It does not cover every cloud feature and is not intended for production operation or security testing.
AWS Machine Learning Blog details an AI-PDLC framework built with six AWS Quick agents across discovery, decision-making, design, and prototyping. The workflow connects enterprise data from tools including Slack, Salesforce, ServiceNow, SharePoint, Jira, Outlook, and email, while producing traceable artifacts and requiring human approval at each gate. AWS says the approach can shorten development cycles, but that productivity benefit remains a vendor claim.
TechNews reports that Amazon shortened the estimated useful life of some servers and networking equipment from six years to five in 2025, citing faster AI and machine-learning progress. The change increased depreciation and amortization by $1.4 billion and reduced net income by $1 billion, with AWS identified as the business most affected. TechNews also cites a forecast that Amazon, Microsoft, Alphabet and Meta will spend a combined $725 billion on capital expenditures in 2026.
AWS has expanded Amazon ElastiCache Serverless Valkey with vector search, full-text search, hybrid search, and server-side aggregation. Applications can combine keyword and vector conditions in one query over continuously updated cached data, with listed use cases including RAG, AI agent memory, product search, and real-time recommendations. The update is available for ElastiCache Serverless caches using Valkey 9 across AWS commercial, GovCloud, and European sovereign cloud regions.
AWS announced that the Web Search API for Amazon Bedrock AgentCore is now available, letting applications and agents retrieve current web results through a single call signed with existing AWS credentials. Results can include titles, URLs, publication dates, and relevant extracts, with request-level domain and date filters, while AgentCore Gateway can apply shared source policies. AWS also detailed IAM controls, SDK support, six launch regions, and pricing of $7 per 1,000 queries.
AWS Machine Learning Blog introduced Amazon Bedrock AgentCore Evaluations, combining AWS-built-in evaluators with DeepEval and AutoEval metrics for measuring AI-agent behavior. The service supports on-demand evaluation for direct session testing, online evaluation of sampled production traffic, and batch evaluation against fixed session datasets, while also offering custom LLM-as-a-judge evaluators. AWS states that its built-in evaluators are quality-tested and benchmarked for reliability, though the evidence comes from the vendor's own tutorial.
Trovata has introduced autonomous treasury agents that run on schedules, triggers, or rules to monitor activity, explain variance, detect anomalies, and deliver updates. AWS Machine Learning Blog says the agents were built with Amazon Bedrock and Amazon Bedrock AgentCore, access treasury data through Trovata’s verified API layer, and support role-based access control.
AWS describes a deployable reference implementation that centralizes Amazon Bedrock model invocation logs from linked accounts into a queryable table. The architecture uses Amazon CloudWatch, Amazon Data Firehose, Amazon S3, AWS Glue and Amazon Athena. According to AWS, the processed data includes prompts, model responses, input and output token counts, model usage and stop reasons, while a separate query derives per-invocation cost. AWS provides two CloudFormation templates, Lambda code and SQL queries, along with architecture details, deployment steps and an end-to-end test script.
Amazon engineers reported that Amazon and AWS teams replaced Claude Sonnet 4.6 with the Strands Decider E4B decision model for high-volume internal ticket labeling. Amazon's measurements showed comparable results, average call time falling from 4,215 ms to 897 ms, and estimated list-price cost dropping about 98%; these figures and the claimed annual savings of millions of dollars were not independently verified.
AWS explains that agents in AWS Quick follow instructions, load task-specific skills, act through connectors, and ground answers in connected spaces, knowledge bases, datasets, and dashboards. The tutorial covers the built-in Quick agent, custom agents, the official skill catalog, and creating and sharing skills. Agents can be used interactively in chat or run through routines, with configurable access to spaces, connectors, skills, and other resources.
Amazon Bedrock Managed Knowledge Base now supports single-shot RAG and retrieval-only requests through AgenticRetrieveStream. Single-shot RAG searches a knowledge base once and generates an answer from the retrieved passages, while retrieval-only requests return matching passages without generating an answer; the existing agentic retrieval mode continues to plan and run multiple searches before answering.
AWS Machine Learning Blog presented the Pluggable Agentic AI Framework as a seven-layer reference architecture for building production-scale multi-agent applications on Amazon Bedrock AgentCore. The layers separate data and knowledge, model inference, orchestration, security and guardrails, application, observability, and business value and cost tracking, with stable contracts that allow underlying technologies to be replaced. AWS also said the architecture is not a substitute for production security hardening, testing, and review.
AWS Machine Learning Blog describes how CrossnoKaye uses Amazon Bedrock AgentCore to ground its AURA agents in live plant data, equipment topology and control logic through the ATLAS platform. AURA runs Python workloads in isolated AWS-managed microVM sandboxes, while edge safety controls and Amazon Bedrock Guardrails constrain its industrial operations. CrossnoKaye reports operating the system across more than 100 facilities and 13,000 controlled assets, though its scale and performance claims have not been independently verified.
AWS announced its first brand campaign for AWS Quick, an AI workspace where teams use agents to research, analyze, and act across existing data and applications. The campaign runs in the United States and United Kingdom through streaming TV, online video, social media, and business publications; AWS says the product can be downloaded immediately, while enterprise pricing is available through an AWS account team.
GPT-6.1 Terra is generally available through Amazon Bedrock, according to a secondary report citing AWS and OpenAI materials. AWS positions Terra as the GPT-6 family’s daily-task model, while vendor claims say its token rates are at least 50% below GPT-6.1 Sol and GPT-5.6 Terra. OpenAI also reports that Terra matched GPT-6 Astra on Agents’ Last Exam and answered more than three times faster, though the benchmark was not independently tested.
This technical guide examines a constructed troubleshooting scenario in which a deleted image remains accessible through an old link or is republished by a late-running generation task. It recommends atomic state checks and conditional commits, private staging objects, idempotent cleanup, and separate verification for access revocation, object removal, CDN invalidation, task convergence, backups, and provider copies. The author ran a local Python state model but did not test real cloud storage, a CDN, a model provider, or a production database.
Amazon SageMaker Inference now supports GPU partitioning that uses NVIDIA Multi-Instance GPU (MIG) to split a physical GPU into independent partitions, each with a fixed share of compute units and dedicated memory isolated in hardware. According to the AWS Machine Learning Blog, this lets customers run more models than an instance has physical accelerators and choose a fraction of a newer accelerator instead of a whole GPU. The feature works on inference component endpoints and is available on P4d, P5, G7 and G7e instance families across seven AWS Regions, subject to instance and quota availability; single-model endpoints, serverless endpoints, heterogeneous instance pools and asynchronous invocation are not supported.
A report on Amazon Quick presents a taxonomy for agentic work organized around four questions: When, What, Who, and How. It defines triggers as mechanisms that start work, agents as actors that interpret instructions and use capabilities, and capabilities as tools, skills, and workflows. It also describes learned context as memory that accumulates through use. The report says most of this framework is designed to remain invisible to end users while giving system builders a shared vocabulary for reasoning about implementation choices.
The AWS Machine Learning Blog describes how an AWS team built a serverless Model Context Protocol (MCP) server on Amazon ECS with AWS Fargate to give AI agents conversational access to an enterprise data lake used by North America Sort Center planning teams. The design puts a metadata catalog, Amazon Bedrock-generated SQL validated for Athena, SELECT-only query tools and Amazon Bedrock Guardrails behind a standardized /mcp endpoint, so any MCP-compatible client can use the same tools. Onboarding a new dataset requires only a JSON metadata file that the server discovers on the next pipeline sync, replacing per-dataset custom integration code.
AWS Machine Learning Blog published Part 2 of a tutorial showing how an AI agent on Amazon Bedrock AgentCore can act for signed-in users through Okta Cross App Access. The implementation routes agents through AgentCore Gateway, evaluates a default-deny Cedar policy set before tool execution, and connects governance decisions to Okta logs through AgentCore Observability and Amazon CloudWatch Omni.
AWS Machine Learning Blog published a guide and sample for running Claude Code as a remote coding agent on Amazon Bedrock AgentCore. Each task runs in an isolated AgentCore Runtime session inside the user’s AWS account and returns a GitHub pull request for human review and merging. The sample also demonstrates private networking, short-lived credentials, least-privilege permissions and Amazon Bedrock data controls.
Atomicwork has launched an Artificial Intelligence Coworker Marketplace offering prebuilt, certified AI coworkers for enterprise IT and security teams. The initial lineup covers enterprise operations, CloudOps, AccessOps, SecOps and DeviceOps, with integrations spanning AWS, Microsoft Azure, Google Cloud Platform, Okta, Active Directory, Splunk and Google SecOps. Atomicwork says administrators can control each agent’s system access, permitted actions and governance, while partners may later contribute skills and integrations.
TrendForce expects global foundry revenue to grow about 30% in 2027, driven by AI and general-purpose server demand, while Chinese foundry capacity may expand by more than 20% annually in 2026 and 2027. It also forecasts that capital spending by nine major cloud service providers will double in 2026 and rise another 60% in 2027 to exceed $1.5 trillion, and projects 36% growth in the AI server market in 2027. Separate forecasts cover AI and consumer MLCC pricing, AR glasses shipments, and humanoid-robot commercialization.
AWS Machine Learning Blog reports that Web Answer is now generally available as a managed Amazon Bedrock AgentCore Gateway tool. It accepts a natural-language question and, according to AWS, plans parallel web searches, verifies facts on pages, and returns a synthesized answer with source citations in seconds. It is available in US East (N. Virginia), priced at $10 per 1,000 tool calls, with a per-account limit of 10 transactions per second.
LlamaIndex announced that LlamaParse can now run in customers’ AWS, Azure, or Google Cloud accounts, or on Kubernetes in their own data centers. Enterprise plans include self-hosting, with deployment teams able to use their own model providers, host supported models on local GPUs, or delegate inference management to LlamaIndex. The vendor says Parse Cost-Effective and Agentic should maintain the same quality with locally hosted document models, while Extract results may vary when deployments use different frontier models.
What problem it solves
Developers using AI coding CLIs need a way to delegate work among multiple agents and coordinate their execution. CAO addresses this by running provider CLIs in isolated terminal sessions and giving a supervisor tools for coordinating workers.
Key features
Multi-agent delegation — A supervisor delegates work to specialist agents, with parallel and sequential execution suppor…
What problem it solves
AI-generated code can violate the rules its developers intended, especially when those rules are described in natural language and the resulting code is not reviewed closely. Bend addresses this by representing important rules as laws and requiring mathematical proofs when the code is edited.
Key features
Proof-enforced application laws — Applications can declare rules in `L…
What problem it solves
Agents can understand diagram content but often produce generic layouts that do not match a site's visual identity. Diagram Design addresses this by applying a defined editorial style to agent-generated diagrams and producing publishable, self-contained files without a manual design workflow.
Key features
Self-contained HTML and SVG output — Each diagram is written as one HT…
What problem it solves
Organizations need a distributed storage layer that can serve objects, files, tables, and vectors while keeping administration and horizontal expansion straightforward. SeaweedFS addresses this by providing these capabilities on shared storage and allowing capacity to grow by adding volume servers.
Key features
Unified distributed storage — A single cluster exposes S3 object…
What problem it solves
Modern systems need controlled access to database credentials, API keys, and other secrets across platform-specific infrastructure. Teams also need secure storage, key rotation, encryption, revocation, and detailed audit logs without building a custom solution.
Key features
Secure secret storage — Vault stores arbitrary key/value pairs and encrypts data before writing it to …
What problem it solves
VPN traffic can expose protocol characteristics that are examined during network inspection. The project addresses this by encrypting UDP tunnels, applying configurable traffic masks, and rotating session keys while providing bounded packet handling and live traffic-quality checks.
Key features
Encrypted and forward-secret sessions — The InitialV2 exchange confirms a fresh s…
What problem it solves
Leaked credentials can be difficult to identify and assess because they may be stored across repositories, chats, wikis, logs, filesystems, and other systems. Finding a credential pattern alone does not establish whether it is active, so the project combines discovery, classification, validation, and analysis.
Key features
Broad secret discovery — Scan sources including Git …
What problem it solves
Creating intelligent AI agents that can collaborate on complex tasks can require code and substantial integration work. Docker Agent addresses this by letting users define agents in YAML, assign tools, and run them through the Docker CLI.
Key features
Multi-agent orchestration — Create teams of specialized agents that delegate tasks automatically.
Source: README
Extensible t…
What problem it solves
AI coding agents need AWS-specific tools, knowledge, and guardrails to work effectively with AWS services. Teams using different coding agents also need a consistent way to configure AWS API access, guidance, and reusable skills.
Key features
AWS MCP Server — The managed MCP Server gives agents access to AWS APIs, isolated script execution, and current documentation through …
What problem it solves
Organizations need to build and maintain custom applications such as dashboards, admin panels, customer 360 tools, IT automations, and service management systems. This platform streamlines those application-development workflows while supporting databases and APIs.
Key features
Low-code custom application development — Build internal applications through a low-code platform …
What problem it solves
Ad-hoc AI coding can lose context as projects grow. AI-DLC keeps requirements, decisions, implementation, tests, and operational work connected through an audited lifecycle.
Key features
Harness-neutral workflow core — A single core runs across multiple coding-assistant harnesses, with thin harness-specific integrations.
Source: README
Structured delivery lifecycle — The wor…
What problem it solves
Teams need to capture, search, and analyze multiple observability signals without managing fragmented tools or complex infrastructure. OpenObserve addresses this with a unified platform, a single-binary deployment option, and familiar SQL and PromQL query languages.
Key features
Unified observability — Logs, metrics, traces, RUM, dashboards, alerts, incidents, pipelines, and…
What problem it solves
Developing against AWS services often requires a cloud account, authentication token, or paid feature gates. Floci provides AWS-shaped services locally so existing AWS SDKs, CLI commands, infrastructure-as-code workflows, and test suites can run without those dependencies.
Key features
Local AWS services without an account — Run AWS-compatible services locally without a clou…