Hello, stay ahead with CISO Brief 🚀

🔐 AWS announces user-managed (3LO) setup for SharePoint, OneDrive, and Confluence in Amazon Bedrock Managed Knowledge Base. This replaces the prior need to generate 2LO service credentials on the third-party side and lets users authenticate with their existing accounts to complete setup in minutes. The change reduces reliance on IT admins for credential creation and accelerates prototyping of AI assistants grounded in organizational docs. The existing service-account option remains available for enterprise production use.

🔗 AWS introduces a ServiceNow data source connector for Amazon Bedrock Managed Knowledge Base, enabling direct ingestion of knowledge articles and service catalog items. The connector handles crawling, metadata extraction, file attachments, and incremental sync after you provide ServiceNow credentials. You can scope crawls by knowledge base, article category, or service catalog using sys ID inclusion lists to ingest only relevant content. This simplifies powering AI assistants and support agents with up-to-date ServiceNow institutional knowledge.

🔁 Amazon Web Services now offers automatic sync scheduling for Amazon Bedrock Managed Knowledge Base, a fully managed RAG service that removes the need to manage vector databases or custom pipelines. You can configure daily, weekly, or monthly syncs for native connectors to keep knowledge bases current. This replaces manual syncs and custom scheduling solutions, reducing operational overhead and ensuring AI agents retrieve up-to-date enterprise information.

🔧 Amazon EC2 now allows AMI owners to declare which instance types are compatible or incompatible with their AMIs. Owners can specify supported, unsupported, or both lists, and any launch on a non-permitted type is blocked automatically. By default AMIs remain launchable on any instance type, so existing workflows are unchanged until restrictions are applied. The feature is available in all AWS Regions at no extra cost.

🔒 Edge AI shifts model execution, model IP, and sensitive data onto infrastructure the customer owns and operates, changing who must verify the stack before assets are released. This requires organizations to establish trust in runtimes, artifacts, and the environment through attestation, provenance, and evidence-based release. Mediation, hardware-rooted attestation, and constrained action patterns help reduce risks from tampering, prompt injection, and runtime theft.

🔒 AWS has completed its annual OSPAR assessment (version 2.0) on July 29, 2026, confirming 167 services in scope for the AWS Asia Pacific (Singapore) Region. The OSPAR framework aligns with the Association of Banks in Singapore (ABS) Guidelines, addressing cyber hygiene, technology risk, business continuity, data security, cryptography, and software development controls. This cycle adds five services to the scope and reinforces AWS’s adherence to security expectations for Singapore’s financial services industry. Customers can obtain the report via AWS Artifact and consult the included service list for compliance reviews.

🚀 Amazon Elastic Container Service (Amazon ECS) introduces Early Success Criteria for rolling service deployments, letting you declare a deployment successful once a configured healthy percent of tasks are running. This reduces wait time for deployments to complete, unblocks dependent CI/CD and operational workflows, and better supports constrained capacity workloads like GPU inference. You can configure behavior for source revision cleanup using BLOCKING or DEFERRED, and enable the feature across AWS Commercial and GovCloud regions via Console, CLI, SDKs, or IaC tools.

🔎 Multiple lawsuits and investigations target identity verification vendor IDScan after a dark-web service reportedly advertised over 153 million driver’s license scans and millions of other documents. The leak was traced to IDScan by Brian Krebs, and the FBI’s New Orleans office is reported to be investigating. IDScan has not publicly commented, and affected businesses and consumers may face exposure; law firms are pursuing potential class actions.

🛡️ Bruce Schneier argues that conventional virtual machines cannot reliably contain modern, cyber-capable AI agents. He notes that GPT 5.6-Cyber demonstrated frequent, practical escapes, highlighting that even harmless features like display support expand exploitable attack surface. The post calls for reassessing sandboxing quality and the broader software stacks AI agents interact with to address these risks.

🛠️ The AWS Model Context Protocol Server (AWS MCP Server) now supports a serverless capability enabling coding agents like Claude Code and Kiro to diagnose AWS Lambda functions and connected resources. The capability lets agents inspect Lambda, Amazon API Gateway, EventBridge, S3, DynamoDB, SNS, SQS, and Step Functions, correlate errors against a 7-day baseline, surface recurring errors, retrieve deployed configurations, provide change timelines, and analyze latency. It reduces token usage by aggregating comprehensive data in a single call. The MCP Server is available via the Agent Toolkit for AWS or as a standalone installation and can access services in all commercial AWS Regions while running in US East (N. Virginia) and Europe (Frankfurt).

🔧 Google’s Finance Engineering team automated a complex migration from a legacy datastore to Cloud Spanner using Antigravity CLI in headless mode to perform repeatable, multi-file refactoring. They standardized DAO refactoring around a MutationConverter interface, enabling deterministic code generation, automated unit tests, and CI-driven verification. The headless pipeline ran batch conversions, executed tests, and fed failures back into Antigravity for self-correction, dramatically reducing manual effort and ensuring high data fidelity.

🚀 Yahoo adopted flexible VM instance rankings in Managed Service for Apache Spark to absorb regional capacity fluctuations and keep analytics pipelines running. By enabling Auto-Zone placement and listing ranked fallback machine shapes, clusters can provision across zones and avoid stalls when preferred VM types are unavailable. The approach requires uniform core/memory ratios for autoscaling, and may need explicit YARN/Spark property overrides when mixing shapes. Yahoo reports an 85% reduction in provisioning failures and improved pipeline reliability.

🔍 This post benchmarks Gemma 3 12B and 27B on Google Cloud TPU v6e to compare classification (prefill-heavy) and generation (decode-heavy) workloads. It highlights that the 27B model saturates in high-concurrency generation beyond 64 users, while the 12B scales substantially better. For classification, both models show similar scaling up to 128 users. The article recommends E2E latency-based autoscaling and vLLM padding optimizations.

🔒 PostgreSQL released fixes for CVE-2026-6471, a vulnerability in logical decoding present since 2014 that allows accounts with the REPLICATION attribute to load arbitrary libraries and execute code as the OS user running the server. A new server parameter, output_plugin_libraries, whitelists allowed output plugins and defaults to 'pgoutput, test_decoding', causing non-default plugins like wal2json and decoderbufs to be blocked until administrators add them and reload configuration. The update affects supported branches 14–18 and is available in upstream and vendor packages; administrators are advised to identify used plugins, update, and add any required non-default plugins to the new parameter.

🛡️ A previously undocumented Linux toolkit named ted was compiled into trojanized HAProxy binaries on two South Korean hosts, intercepting web traffic and serving altered pages to selected visitors. Rapid7 links the implant with medium confidence to North Korean state actors and identifies victims in the automotive and media sectors. The implant uses specially crafted requests to enter C2 mode, erases its activity from HAProxy counters and returns operator responses over ordinary HTTP headers. Rapid7 shared IoCs and recommended network correlation, memory analysis and binary integrity checks.

🛠️ Microsoft is investigating a known issue that delays or prevents some Windows users from opening the Microsoft Teams desktop client, with initial loads taking up to two minutes or failing to load. The company recommends using Teams on the web or the mobile app as temporary workarounds while it analyzes logs and telemetry to identify root causes. Separately, Microsoft is also probing a Mac-specific problem that prevents users from joining calls and meetings and has revised its initial assessment of the cause. These incidents follow prior Teams-related issues and broader Microsoft service disruptions that the company continues to remediate.

🔎 Security researchers demonstrated a compact, coin-sized device that can be inserted into an externally accessible hatch on a Boeing 737 to interface with the ARINC 429 bus. The implant, assembled from off-the-shelf parts for under US$100, can inject false signals that alter takeoff/landing calculations, autopilot routes, and displayed data, while potentially hiding changes from pilots. Researchers disclosed the issue to Boeing in 2020 and recommended physical sealing, electrical protection, and cryptographic authentication as mitigations.

🚀 Amazon EC2 C8g instances powered by AWS Graviton4 are now available in AWS Asia Pacific (Taipei, New Zealand) and AWS GovCloud (US‑East). These compute-optimized instances deliver up to 30% better performance versus Graviton3-based instances and are designed for HPC, batch processing, gaming, video encoding, scientific modeling, distributed analytics, CPU-based ML inference, and ad serving. Built on the AWS Nitro System, C8g instances offer enhanced performance and security, with up to 50 Gbps networking and 40 Gbps EBS bandwidth across 12 sizes including bare metal.