Firecracker
Firecracker Escape Flaw Puts Cloud Multi-Tenant Isolation at Risk
A newly disclosed Firecracker VM escape could undermine tenant isolation as AI services increasingly rely on microVMs to run models and tools.
Back to home
AWS
30 articles tagged with this topic
AWS
Deleted AI Images Can Return, Exposing a Data Consistency Failure
A runnable script reveals how concurrent generation and deletion can resurrect AI images, and why databases and object storage often fall out of sync.
Amazon Quick
Amazon Breaks Agents Into Unified Building Blocks for Easier Buying
Amazon Quick defines six agent primitives, pushing enterprise buying toward task-, permission-, and reliability-based systems.
Amazon SageMaker
SageMaker Opens Up GPU Partitioning as AWS Tackles Idle Compute
AWS added NVIDIA MIG support to SageMaker Inference, splitting one physical GPU into up to 7 isolated instances with flexible allocations.
Amazon Bedrock AgentCore
Amazon Packages Web Search Into an Answer Service, Making AI Agents Deliverable
AWS launches Web Answer for Bedrock AgentCore, combining search, verification, and cited answers in one managed API call.
Amazon Bedrock
AWS Makes Agent Evaluation a Quantifiable Decision: Enterprise Adoption Enters Fine-Grained
AWS adds decision models to AgentCore Evaluations, enabling calibrated probabilities, risk screening, and deeper LLM review before agents go live.
U Mobile
U Mobile Lets AI Run Telecom Tickets as Enterprise Agents Hit Core Operations
U Mobile's AiDaH now spans 150+ use cases, cutting some critical incident tickets by 84.5% and signaling a shift from AI assistance to execution.
Indegene
Indegene Deploys Data Quality Agents as AI’s Bottleneck Shifts to Trust
Indegene built a data quality agent on AWS that generates, validates, and operationalizes rules for life sciences.
AWS
AWS’s Four-Layer Method for Production-Ready AI Agents
AWS outlines a four-layer context engineering method for production AI agents, cutting context use by up to 95% through on-demand skills and written e
AWS
AWS Makes GPUs Shareable: Compute Is Becoming Infrastructure
AWS's multi-tenant SageMaker HyperPod enables GPU sharing, isolation and chargeback. We see compute becoming infrastructure managed like a utility.
AWS
AWS Exposes the Permission Problem Enterprise AI Has Been Dodging
Wiring AI into internal docs risks leaking financials or strategy across permission gaps. AWS's solution this week admits the problem is far harder th
Claude Haiku 5.5
Claude Haiku 5.5 is 75% Cheaper — LLMs Go Pay-Per-Use
Claude Haiku 5.5 lands on Amazon Bedrock, ~75% cheaper than its predecessor. The cost curve hasn't bottomed — pay-per-call is now the default for ente
AWS
AI Assistants Escape 'Goldfish Memory' — AWS Makes 'Remembering You' Standard
AWS open-sources a template combining OpenClaw with AgentCore to give AI persistent memory—signaling memory becoming the new industry standard.
AWS
AWS Bedrock Adds Agentic Retrieval — Compute-Heavy Overkill for Most Use Cases
AWS Bedrock adds Agentic Retrieval: AI decomposes queries and re-searches when evidence is thin. Upgrade or over-engineering for most enterprises?
AWS
AWS Makes Agent Cross-Account Migration Hands-Free — Enterprise AI's Next Hurdle
AWS uses an MCP server to automatically move Agent resources between dev and production accounts. Not sexy, but it signals enterprise AI deployment is
AWS
AWS Packs Cloud Deployment Into Coding AI — Barriers Drop, Lock-In Deepens
AWS ships SageMaker as an MCP plugin for Claude Code, Kiro, and Codex, letting coding AI Agents handle deployment and tuning. Cloud vendors compete on
Anthropic
Claude Code Clears Pentagon's Top Bar — Compliance AI Finally Gets Real
Claude Code, Opus 5.5 and Sonnet 5.5 clear FedRAMP and DoD IL4/IL5 on AWS GovCloud — opening AI coding assistants to ITAR-bound projects.
turbopuffer
A Vector Database Vendor Says 'RIP' — RAG's Foundation Is Being Quietly Rewritten
Vector databases powered RAG for two years. Now turbopuffer itself posts 'RIP, vector database.' Does enterprise AI follow?
Firecracker
AWS's decade-old Firecracker makes a comeback as the AI Agent sandbox
Firecracker, the microVM behind AWS Lambda, was dormant for years. The AI Agent industry is reviving it as a code sandbox—containers are no longer con
Firecracker
The Real Reason AI Agents Stall in Production: Infrastructure, Not Intelligence
Firecracker engineering docs reveal AI agents fail mid-task from infrastructure gaps, not weak models: lost state, duplicate actions, broken cleanup.
AWS
AWS Brings AI Apps Live: The Real Win Is Permission Isolation
AWS added live data to its AI app product: each launch re-queries the database, filtering by user identity. Two production blockers solved: data fresh
AWS
AWS Cuts Migration Code from 4 Weeks to Minutes — The Toolchain Bar Is High
AWS uses four AI Agents to cut 300+ app code generation from 3-4 weeks to minutes — but the MCP toolchain requirement is steep.
AWS
Amazon Bandit Ad Test: 7 Weeks Prove Bad Content Beats Smart Algorithms
Amazon Payments' 7-week contextual bandit test boosted conversions for one user segment only. The verdict: bad creative beats clever algorithms.
Amazon Nova
Uniopen hands Amazon AI 9 retail behaviors — copy the pipeline, not the model
Uniopen fine-tuned Amazon Nova 2 Lite for 9 user behaviors × 3 object types. The lesson: copy the data-train-review-deploy pipeline, not the model.
AWS
AWS 'How to Ask AI' Guide: Enterprise AI's Bottleneck Is Users, Not Models
AWS published a prompt engineering guide for Amazon Quick. The signal: AWS admits the bottleneck in enterprise AI is users, not models.
xAI
xAI Lands Grok 4.7 on AWS Bedrock: Musk Finally Cracks Enterprise AI
xAI lands Grok 4.7 on AWS Bedrock — 500K context and self-verification are the tech; cracking enterprise AI distribution is the real story.
Anthropic
Claude Sonnet 5.5 Lands on AWS — Anthropic's Smaller-Model Enterprise Bet
Anthropic puts cheaper, faster Sonnet 5.5 on AWS Bedrock. AWS enterprises get the latest Claude without cloud switching — clear signal: trading distri
AWS
AWS Uses AI to Test Web Pages for You — The 15-Year Selenium Era Begins to Crack
AWS turns browser AI agents into a cloud service: natural language replaces ten lines of Selenium, freeing synthetic monitoring from fragile scripts.
AWS
AWS Packages AI Image and Video as a Service — vLLM-Omni Catches Closed APIs
AWS shipped a SageMaker tutorial running FLUX.2 + Wan2.1 via vLLM-Omni—self-hosted multimodal generation just got cheaper.
AWS
Qwen3-TTS Lands on AWS SageMaker — Open-Source Voice Models Go Cross-Platform
Qwen3-TTS appears in official AWS SageMaker tutorials as a streaming voice output demo. Open-source multimodal models gain their cross-cloud entry.