Amazon Bedrock
AI overview
Sign in and the AI will write an overview from our coverage.
Headlines · 71
- A Practical Guide to Managing Data Before Using AI
少数派 AI argues that users should understand where their data goes and how they can export or retain it before focusing on advanced prompt techniques. Drawing on a Notion sync-loss incident and platform privacy policies, it recommends disabling training where possible, using temporary chats for sensitive material, maintaining local copies, and avoiding plaintext credentials in AI contexts.
少数派 AI · 🔥 9 - Chinese Edition of Anthropic’s “Building Effective Agents”
A Chinese translation and illustrated edition of Anthropic’s “Building Effective Agents,” credited to original authors Erik Schluntz and Barry Zhang and dated December 19, 2024. The report describes it as a complete translation with seven redrawn architecture diagrams. It distinguishes workflows, which follow predefined code paths, from agents, which dynamically direct their own processes and tool use. The guide covers prompt chaining, routing, parallelization, orchestrator-workers, and evaluator-optimizer patterns, and recommends starting with direct LLM API calls, keeping designs simple and transparent, and carefully designing tool interfaces. The supplied report does not independently verify the translation’s completeness or editorial fidelity.
掘金 · 人工智能 · 🔥 4 - Baseten Grid enters public preview for enterprise AI routing
Baseten introduced Grid in public preview as an enterprise control layer for routing AI requests, monitoring model usage, managing budgets and access, and applying centralized guardrails. It combines Baseten Router, a gateway supporting providers including OpenAI, Anthropic, Gemini Enterprise Agent Platform, Amazon Bedrock, and Baseten, and configurable safety policies. Baseten says its early benchmarks found nearly 70% lower cost per coding task with a 0.9-percentage-point quality tradeoff versus Opus 5.5, though results depend on workload and desired tradeoffs.
Baseten Blog · 🔥 5 - Anthropic expands Claude watermarking with sampling-setting caveat
Anthropic told API customers that text watermarking will be retrofitted to Claude Opus 4.6, Claude Opus 4.5, Claude Sonnet 4.5, and Claude Haiku 4.5 starting November 16, 2026; the rollout on Amazon Bedrock may take several additional days. Claude Opus 4.6 was initially expected to receive the watermark on November 3, but Anthropic confirmed the delay without explaining it. The report says all four models in this wave apply the watermark only when the sampling parameters temperature, top_p, and top_k remain at their defaults. Anthropic says a minor change to any one parameter is enough to avoid the watermark on Claude Opus 4.6 and Claude Sonnet 4.6, and warns developers who rely on watermarking for compliance to keep those settings unchanged. Anthropic says the watermark adds no tokens, does not affect pricing or request and response formats, and contains no information about users, their organizations, or their conversations.
The Decoder · 🔥 5 - Interconnects launches free dashboard for open-model adoption
Interconnects has launched the Interconnects Dashboard, a freely available aggregator of open-model adoption and performance data gathered from across the web. Its accompanying research uses Baseten, Fireworks, Together AI, and OpenRouter as reference points to estimate the open-model inference market, including Fireworks’ reported service of more than 40 trillion tokens per day. Interconnects cautions that platform metrics, startup marketing figures, and public evaluations are noisy or incomplete, making its market estimates preliminary.
Interconnects · 🔥 5 - AWS details four layers of context engineering for production AI agents
AWS Machine Learning Blog published a working-code guide to four context-engineering layers for production AI agents: architectural isolation, prompt control, filesystem externalization, and independent verification. The guide applies these practices in the open-source Deep Insight multi-agent data-analysis solution, built with Strands Agents SDK and Amazon Bedrock AgentCore. AWS reports that its write-and-execute tool reduced script-added context tokens by about 95%, while on-demand skill loading reduced the system prompt by about 97%.
AWS Machine Learning Blog · 🔥 5 - Indegene deploys LLM-powered data quality agent on AWS
The AWS Machine Learning Blog reports that Indegene built a production-grade, LLM-powered Data Quality Agent within the NEXT Unified Data Platform on AWS. The agent generates context-aware data quality rules, validates them, and produces ready-to-run SQL and Python under centralized, human-in-the-loop governance. Indegene estimates that anomaly detection is 80–85% faster and rule maintenance effort is 85% lower in its production deployments.
AWS Machine Learning Blog · 🔥 5 - OpenAI says India has overtaken the US in ChatGPT weekly users
OpenAI India managing director Prabhjeet Singh told The Economic Times that India now has more ChatGPT weekly active users than any other country, including the United States, and that its user count has doubled since January. He also called India ChatGPT's largest consumer market, though he did not disclose a current user total and the claim was not independently verified. IT之家 reported that OpenAI previously said in February that India had more than 100 million weekly active users and was its second-largest market after the United States.
IT之家 AI · 🔥 8 - GPT-5.6 Cyber now available via Daybreak Red in AWS GovCloud
OpenAI announced that GPT-5.6 Cyber is now available to approved customers in AWS GovCloud (US-East) through Daybreak Red. OpenAI says the purpose-trained cybersecurity models support authorized vulnerability research, exploit validation, security testing, and mitigation development, with qualified security practitioners expected to review findings and proposed fixes before implementation. Access requires both Daybreak Red approval and AWS provisioning.
OpenAI · 🔥 5 - Mandiant details LLMjacking abuse of stolen cloud credentials
Mandiant investigated multiple LLMjacking incidents in 2026 in which attackers used stolen cloud credentials and API keys to make unauthorized, high-volume use of enterprise AI services. In one case, an attacker routed stolen API access through Jet Night and made more than 500,000 LLM API calls in under eight hours, leaving the victim organization with nearly US$1 million in usage costs. Other attackers activated services such as Amazon Bedrock and Gemini API inside compromised cloud environments, while some resold access through the illegal proxy operation LLM Mixers.
iThome 台湾 · 🔥 3 - Gen Digital details Norton Revamp’s AWS-powered AI news engine
AWS Machine Learning Blog says Gen Digital launched Norton Revamp in March 2026, a generative AI platform for professional reputation management built by a small engineering team. Its News Engine uses Amazon Bedrock AgentCore Web Search to ground content suggestions, goal recommendations, and personalized feeds in current web results with source URLs and publication dates. Gen Digital and AWS say the managed service reduced infrastructure overhead and data-egress risk, while the platform also uses Amazon Nova Lite, Claude, Stability AI, and Amazon Bedrock.
AWS Machine Learning Blog · 🔥 3 - Amazon Bedrock Adds India-Based Inference for Kimi K3
Amazon Web Services announced that Amazon Bedrock now supports in-country inference for Kimi K3 through an India geographic profile, restricting processing to the Mumbai and Hyderabad Regions. The model remains available through global and US-based cross-Region inference profiles, with AWS saying requests can originate from a wide range of source Regions. AWS also described multimodal support for images, PDFs, tool calling, structured outputs, and prompt caching for Indic-language document workflows.
AWS Machine Learning Blog · 🔥 3 - Enterprise AI Moves Into Operations, Pricing and Governance
A report published by 智讯智库, drawing on public financial reports and company disclosures, identifies six enterprise-AI trends: AI-related spending and delivery are scaling; applications extend from coding into procurement, marketing, customer service and transactions; measurable efficiency gains remain distinct from unproven company-wide revenue and profit gains; pricing is shifting beyond headcount and seats toward usage or business outcomes; multi-model and hybrid deployment are becoming more common; and agents that can execute actions require stronger permissions, monitoring, auditing and human handoffs. Infosys, TCS and Accenture also use different definitions for AI-related revenue, limiting direct comparisons.
虎嗅 AI · 🔥 0 - CrowdStrike to Acquire Fiddler AI and Dreadnode
CrowdStrike signed agreements to acquire Fiddler AI, an observability platform for monitoring AI agents, and Dreadnode, which tests AI models, applications and agents for weaknesses. Fiddler's technology will become Falcon Observe in the Falcon platform, while Dreadnode's automated testing capabilities will also be integrated into Falcon; both deals and the related products are expected in CrowdStrike's fourth fiscal quarter of 2027, subject to customary conditions. The Wall Street Journal reported values of about $400 million for Fiddler AI and $95 million for Dreadnode, but CrowdStrike did not disclose the prices.
Techmeme · 🔥 6 - AWS tutorial: distill a shopping agent with sequence-level knowledge distillation
The AWS Machine Learning Blog shows how to build a lower-cost shopping agent on Amazon SageMaker AI using sequence-level knowledge distillation (SeqKD), generating multi-step trajectories with a frontier teacher on Bedrock, keeping only those that solved the task with ShoppingBench's scorer, and fine-tuning a smaller student on serverless infrastructure. The post also covers encoding trajectories as chat-formatted messages, tracking training with MLflow, gating quality with SageMaker AI model evaluation and a custom tool-selection scorer, and serving the student behind an endpoint that runs the tool loop. AWS reports the payoff was a 4B student that recovers essentially all of its frontier teacher's task quality.
AWS Machine Learning Blog · 🔥 0 - Circle K Cuts DevOps Triage Time with Amazon Bedrock Agents
Circle K built an agentic AI system on Amazon Bedrock that investigates DevOps incidents and sends structured summaries of probable causes to Slack within minutes. According to the AWS Machine Learning Blog and Circle K’s internal incident-response metrics, mean triage time fell from 45–60 minutes to under 15 minutes, a reported reduction of 67–75%. The system initially used Amazon Bedrock Flows and Knowledge Bases before moving to a single agent on Amazon Bedrock AgentCore Runtime.
AWS Machine Learning Blog · 🔥 0 - Northflank Breach Exposes Hosted AI API Keys
Northflank confirmed that some AI API keys hosted on its platform were used without authorization and required customers to rotate sensitive credentials. Its preliminary investigation focused on sensitive data in the control plane, while the company said it had found no evidence that attackers accessed running customer workloads or related data. Wiz tracked two separate cases involving stolen credentials and large-scale Amazon Bedrock API calls, including about 800,000 calls in 30 hours and 1.7 million in 13 hours.
iThome 台湾 · 🔥 6 - Anthropic expands Claude watermarking despite a sampling loophole
The Decoder reports that Anthropic plans to add text watermarking to Claude Mythos 5, Claude Opus 4.7, Claude Opus 4.6, and Claude Sonnet 4.6, while Claude Opus 5.5 already includes the feature. On Opus 4.6 and Sonnet 4.6, the watermark is applied only when temperature, top_p, and top_k remain at their defaults, allowing API users to avoid it by changing any one setting. Anthropic describes the watermark as an imperceptible word-choice pattern that adds no tokens, leaves pricing and API formats unchanged, and barely reduces output speed, but the supplied report contains no independent measurements of these claims.
The Decoder · 🔥 0 - Claude Code v2.1.316 Adds Gateway Controls and Model Settings
Claude Code v2.1.316 added shared monthly group budgets, layered managed policies, and guardrail input tagging for Amazon Bedrock upstreams in the Claude apps gateway. It also introduced controls for disabling the ultrathink keyword, retaining looped cross-certificates, and setting subagent effort, alongside fixes for shell execution, model fallback, restricted file access, and other issues. The release notes also deprecate fast mode for Opus 4.8 on October 30, 2026, with removal scheduled for November 30, 2026.
Claude Code Releases · 🔥 0 - Five-Step Playbook for Managing Amazon Bedrock Costs
A cost-management playbook for Amazon Bedrock organizes spending controls into five stages: attribution, observability, budget enforcement, platform optimization, and recurring review. It recommends assigning every dollar to an identity, team, and application; logging requests with sufficient metadata; using alerts or blocks according to overspend risk; and managing model routing, caching, and token limits centrally. The playbook also recommends a scorecard focused on cost per successful outcome and using the AWS Pricing Calculator, Amazon Bedrock pricing information, and workload-level measurements rather than published averages.
AWS Machine Learning Blog · 🔥 0 - Worldwide Express deploys Bedrock agents for freight billing exceptions
Worldwide Express, working with AWS and LTM, built a multi-agent AI solution using Amazon Bedrock that connects to its SpeedShip transportation management system. The system routes billing exceptions to specialized agents for invoice matching, pricing-variance analysis, and accessorial-charge validation. Before making recommendations, the agents retrieve carrier-specific billing rules, while cost finalization and an AI-generated explanation go to a human reviewer in the SpeedShip analyst interface. Six months after launch, Worldwide Express reported deployment across more than 40 LTL carriers, weekly closure of about 10,000 exceptions with an error rate below 1%, a 75% reduction in active exceptions, a four-week improvement in time-to-cash, more than $1 million in annual savings and a one-time $10 million free-cash-flow release. The AWS Machine Learning Blog attributes these figures to the company.
AWS Machine Learning Blog · 🔥 0 - ChatGPT desktop app now available via Bedrock in AWS GovCloud
The ChatGPT desktop app is now available on Amazon Bedrock in AWS GovCloud (US), according to the supplied report. It provides Work for complex, multi-step tasks and Codex for software development, testing and code review. The described setup routes inference through the Responses API and Amazon Bedrock Mantle in us-gov-west-1, with access governed by AWS IAM and existing GovCloud network controls. The report also says processing can remain within the selected Bedrock region, billing is per token without seat licenses or per-developer commitments, and a configurable zero-data-retention mode prevents prompts and responses from being written to durable storage by AWS or shared with model providers. Organizations remain responsible for reviewing and authorizing their workloads.
AWS Machine Learning Blog · 🔥 0 - Claude Code v2.1.314 Adds Write Controls and Gateway Features
Claude Code v2.1.314 adds a setting that blocks Write, Edit and NotebookEdit operations outside working directories in every permission mode. It also introduces IdP group-based routing for Claude apps gateway upstreams, Bedrock guardrail tracing, and idle compaction for conversations exceeding 200K tokens in VS Code and Claude Desktop's Code tab. The release also addresses effort-level errors, startup failures, image-heavy Bedrock sessions, and Microsoft Foundry deployment fallbacks.
Claude Code Releases · 🔥 0 - Amazon AWS Growth and Heavy AI Investment Coexist
Amazon reported Q3 2026 revenue of $201.94 billion. AWS revenue rose 39% to $45.98 billion and operating income increased 55% to $17.77 billion, exceeding the analyst estimates cited by CNBC. AWS represented about 23% of Amazon’s revenue but roughly 63% of its operating income, according to iThome. Amazon said AI spending drove most of the decline in trailing-12-month free cash flow, which fell to negative $42.4 billion from positive $14.8 billion a year earlier. The company maintained an estimated $220 billion in cash capital expenditures for 2026, primarily for AI and cloud infrastructure.
iThome 台湾 · 🔥 0 - Amazon Quick's UK user group expands through three 2026 meetups
AWS reports that the UK Amazon Quick User Group held three quarterly meetups in 2026, drawing more than 150 attendees across the events. Participants worked with Amazon Quick tools for custom agents, app creation, automation, research, and data workflows, while customer examples covered incident management, governed analytics, and AI-assisted support. AWS also says the Amazon Quick Community provides more than 600 learning resources.
AWS Machine Learning Blog · 🔥 0 - Codex CLI 0.166.0 adds transcript search, Terra support, and diagnostics
OpenAI Codex CLI Releases reports that version 0.166.0 adds keyboard-driven transcript selection with word, line, and page motions, integrated search, and configurable shortcuts. The release also adds GPT-6.1 Terra to the bundled, Vertex AI, and Amazon Bedrock model catalogs, introduces `codex queue --resume`, improves safety-block explanations, and provides focused network diagnostics through `codex doctor --network-only`. Fixes cover thread resumption and forking, app-server lifecycle races, TUI reconnection, macOS cleanup, and reasoning-token accounting.
OpenAI Codex CLI Releases · 🔥 0 - AWS details enterprise Agentic AI Platform architecture on Bedrock AgentCore
The AWS Machine Learning Blog published a walkthrough of an enterprise Agentic AI Platform built on Amazon Bedrock and Amazon Bedrock AgentCore, covering a multi-account operating model, a central Platform control plane, repeatable team-owned Workstream cells, separate inference and MCP tool gateways, and an evaluation pipeline that gates agents before production promotion. AgentCore is described as providing modular runtime, gateway, identity, memory, policy, registry, and evaluation services, while a companion blueprint adds ownership, account boundaries, software delivery, observability, cost governance, and retirement as infrastructure as code. The post states the deployment and available features are based on the eu-west-1 (Ireland) Region and that the repository is a reference implementation developers can adjust.
AWS Machine Learning Blog · 🔥 0 - How to Handle Claude 529 Overloaded Errors
OpenRouter says a Claude 529 overloaded_error indicates that Anthropic’s API temporarily lacks capacity for the requested model, rather than a problem with the user’s key, plan, or request. It recommends retrying with exponential backoff and randomness, then routing the same Claude model to another provider; Anthropic SDKs and Claude Code already perform such retries. OpenRouter’s overload example used a local API stand-in because a real 529 cannot be triggered on demand.
OpenRouter Blog · 🔥 0 - Property Finder Finds Bias in Vision Model Evaluation
Property Finder discovered that 96.8% of its evaluation labels came directly from predictions by its incumbent vision model rather than independent human verification. In an AWS Machine Learning Blog case study, the team rebuilt expert-annotated ground truth and used evolutionary prompt optimization on Amazon Bedrock to select a model with higher accuracy and 86% lower cost.
AWS Machine Learning Blog · 🔥 0 - Exowatt builds reusable SageMaker fine-tuning pipeline
AWS Machine Learning Blog reports that Exowatt used Amazon SageMaker AI serverless model customization to create a repeatable pipeline for task-specific fine-tuned models. A fine-tuned Qwen3-1.7B handled offline alarm triage, while a fine-tuned Qwen3-14B supported cloud energy-market analysis; AWS claims higher accuracy than frontier models at up to 49× lower cost, though those results were not independently verified here. The pipeline turns operational data into a certified training set, evaluated model and deployment artifact, reducing Exowatt's reported workflow from months to days.
AWS Machine Learning Blog · 🔥 0
Experience and discussion from the community
Share my Amazon Bedrock experienceAsk about Amazon Bedrock
Nobody has shared their experience with Amazon Bedrock yet.