Volume 343, Issue 2 · 6 articles from 3 issues
AWS's Deception Benchmark shows AI models catch most real vulnerabilities but flag 41-99% of safe code.
DevOps.com reports that AWS built the Deception Benchmark, containing 14,822 adversarially hardened samples in 16 languages across more than 70 CWE categories, after evaluating 12 models from five providers on whether AI can tell exploitable vulnerabilities from code that only looks risky.
Models identified up to 95% of real vulnerabilities but flagged 41% to 99% of safe code, and proof-of-exploit prompting cut false positives yet missed 7% to 44% of real flaws.
Environment-gated challenges, such as checking whether a Kubernetes network policy blocks an SSRF path, proved hardest, and no configuration kept both error types below 10%.
AWS previews Well-Architected Agent, an AI service that analyzes environments and delivers goal-aligned fixes across 65+ services.
AWS announced a public preview of AWS Well-Architected Agent, an AI-powered service that analyzes an AWS environment and provides contextual recommendations to improve cost, security, performance and resilience.
It correlates utilization metrics, resource configurations and application topology against Well-Architected best practices across more than 65 AWS services.
Recommendations are aligned to declared business goals, come with ready-to-implement fix packages and surface cross-pillar trade-offs, replacing manual audits and generic checklists.
AWS's weekly roundup covers GPT-6 Sol and Luna and Claude Opus 5.5 arriving on Amazon Bedrock, plus Strands harness and more.
The AWS Weekly Roundup for September 28, 2026 centers on model choice on Amazon Bedrock, where OpenAI's GPT-6 Sol, aimed at demanding development and operations work, and GPT-6 Luna, for high-volume repeatable tasks, arrived at lower pricing than GPT-5.6.
Anthropic's Claude Opus 5.5, the first of the Claude 5.5 family, does more with fewer tokens and is tuned for agentic coding and long-running tasks.
The roundup also covers the Strands harness and observability launches for agentic workloads.
AWS weekly roundup covers Bedrock Managed Agents powered by OpenAI, new frontier models, and a Well-Architected Agent preview.
This week's AWS roundup highlights the public preview of Amazon Bedrock Managed Agents powered by OpenAI, a customized AWS-native version of OpenAI's Agents API that runs inside customers' accounts with existing identity and governance controls.
New Bedrock models include OpenAI GPT-6.1 Sol, a GPT-6 Astra UltraFast speed tier, Anthropic Claude Sonnet 5.5 and SpaceXAI Grok 4.7, alongside a preview AWS Well-Architected Agent offering cost, security, performance and resilience recommendations.
Aurora PostgreSQL can now query Apache Iceberg and Parquet data directly, and S3 Tables support all Apache Iceberg V3 data types.
AWS's weekly roundup covers Prime Day 2026 infrastructure metrics, Claude Haiku 5.5 on Bedrock and Amazon Nova 2.5 Sonic.
The AWS News Blog weekly roundup for October 13, 2026 leads with the Prime Day 2026 metrics, including Amazon ElastiCache peaking at 2.3 quadrillion daily requests and DynamoDB processing tens of trillions of calls.
It notes that Claude Haiku 5.5 arrived on Amazon Bedrock at roughly 75 percent less than Haiku 4.5, Amazon Nova 2.5 Sonic made voice agents faster, and Strands Box added boundaries around what agents are allowed to do.
The post then lists the week's other AWS launches and news.
AWS Lambda Web Functions run standard HTTP web apps on Lambda without rewriting them as event handlers.
AWS announced Lambda Web Functions, a new Lambda form factor for running standard port-listening HTTP applications built with frameworks such as Express, Hono, Fastify or Next.js.
Applications deploy with a single command, and Lambda handles scaling, multi-region distribution and active CPU billing without adapters or framework lock-in.
Lambda now offers three form factors on the same Firecracker isolation technology, and existing functions are renamed Lambda Event Functions to reflect their event-driven handler model.