Hello, stay ahead with CISO Brief 🚀

🛡️ Microsoft tracked a macOS ClickFix operation using over 250 front-end domains that fingerprint visitors before deciding whether to show a malware lure. The server-side gate hides malicious pages from crawlers and sandboxes while showing selected Mac users a fake download that ultimately retrieves scripts to launch infostealers such as MacSync and Atomic Stealer (AMOS). The attack still requires users to paste and run an obfuscated Terminal command, and Microsoft recommends users never paste browser instructions into Terminal.

🛡️ OpenAI says it dismantled a Poipet-based scam operation that used ChatGPT to run investment, romance, gambling, and law-enforcement impersonation schemes. The company banned a coordinated cluster of accounts tied to Poipet that created fake personas, generated promotional content, translated messages, and handled administrative tasks. OpenAI investigated in partnership with WhatsApp and highlighted the hybrid, opportunistic nature of modern scam networks.

🛡️ Proofpoint has identified a phishing campaign impersonating COLDCARD that lures victims with a bogus "Hardware audit" notice tied to a recent wallet vulnerability and large Bitcoin theft. The scam directs targets to a clone site that downloads a batch file which escalates privileges, decodes embedded files, and installs a signed decoy plus a ConnectWise ScreenConnect remote access tool. Once connected to the actor-controlled ScreenConnect server, attackers can remotely access systems, steal data or cryptocurrency, and deploy additional malware or ransomware.

🔒 Today Amazon Cognito launches provisioned limits in the console to let teams self-service authentication rate adjustments in minutes. The feature separates an account-level maximum (managed via Service Quotas) from a provisioned limit you pay for and control in the Amazon Cognito console, enabling rapid scaling for events like Black Friday. It supports granular RPS adjustments, programmatic APIs, and cost optimization by billing only for provisioned capacity above defaults.

🔍 AI Insights in AWS Marketplace explains product pricing directly on listings, showing what pricing units map to, how costs scale with usage, and how multiple pricing dimensions combine. It cites sources drawn from the seller's Marketplace listing and public website so buyers can verify explanations. The feature is live in most listings across commercial AWS Regions and sellers can request edits via the Contact Us link.

🆕 Amazon Web Services announces general availability of native vector search for Amazon DynamoDB. The feature enables indexing and approximate nearest neighbor searches over embeddings with single-digit millisecond latency and 99%+ recall at any scale, including trillions of vectors. You can store embeddings alongside other attributes, choose a model to generate vectors (including Amazon Bedrock models), create vector indexes, and apply attribute filters. DynamoDB vector search preserves serverless benefits—no infrastructure management, zero downtime, and pay-as-you-go—supporting use cases such as agent memory retrieval, semantic search, recommendations, and personalized advertising.

🔒 KuppingerCole’s 2026 Leadership Compass identifies a shift in CNAPPs toward unified platforms that secure cloud and AI workloads. The report names Microsoft a Leader across Overall, Product, Innovation, and Market for Defender for Cloud, citing its integrated approach to cloud, data, identity, and AI risk. The post highlights risk-based attack path analysis, AI security posture management, and agentic AI support for detection, prioritization, and remediation.

🚀 UiPath re-architected its infrastructure to support agentic AI and high-scale intelligent document processing by moving from isolated clusters to a shared Google Cloud GPU fleet. The company balances A3 (NVIDIA H100) instances for training with G4 (NVIDIA RTX Pro 6000) instances for inference, using Google Cloud AI Hypercomputer and Dynamic Workload Scheduler to secure predictable capacity. This shared-fleet approach maximizes utilization, reduces costs, and lets engineering teams focus on model performance rather than infrastructure.

🔎 This article explains how shifting from a monolithic data pipeline to a sharded hub-and-spoke architecture reduces the impact of "noisy neighbor" tenants. The Hub acts as a lightweight router while Spokes provide isolated processing with Pub/Sub buffers between them. The design enables independent scaling, fault isolation, tiered pipelines for priority tenants, and spoke-level best practices such as DLQs, strict connection pooling, and asynchronous I/O.

🚨 The U.S. Cybersecurity and Infrastructure Security Agency has issued an urgent directive requiring federal agencies to mitigate actively exploited vulnerabilities in IBM Langflow, N-central, and Apache Tomcat within three days. The most severe issue, tracked as CVE-2026-9198, impacts Langflow and permits unauthenticated remote code execution via chained API endpoints. Vendors have released patches and a hotfix, but incomplete fixes and public proof-of-concept exploits have enabled ongoing attacks. CISA added all three flaws to its Known Exploited Vulnerabilities catalog and urged immediate remediation.

🛡️ Microsoft Threat Intelligence tracked a macOS ClickFix campaign distributing infostealers such as MacSync and Atomic Stealer (AMOS) through a large family of look-alike domains. The operation shifted from embedding the malicious ClickFix lure in page HTML to hiding it behind a server-side browser-fingerprinting gate that selectively shows the lure only to visitors resembling genuine macOS browsers. The blog describes domain patterns, fingerprinting checks, infection chain, detection coverage, and hunting pivots defenders can use to find related activity.

🛡️ Two critical vulnerabilities in the open-source AI control plane Paperclip allow attackers to import a malicious agent and trigger command execution on either network-accessible servers or local developer machines; a third flaw exposes sensitive control-plane details via inadequately guarded API routes. Vendors have released fixes in the source tagged v2026.416.0, which enforces stricter import permissions and hostname validation, and operators are urged to upgrade and review deployment exposure.

🚀 Amazon Aurora Serverless now provides a faster initial scale-up, delivering up to 12 ACUs within one second and continuing to scale up to 256 ACUs as workloads grow. This capability, enabled by default on platform versions 3 and 4, is well suited for agentic AI and bursty applications because clusters automatically scale down to zero when idle. Existing clusters on versions 1 and 2 can upgrade to platform version 4 to take advantage of the improvement. Check platform version in the AWS Console or via the RDS API to confirm eligibility.

🔒 New research from Oasis Security disclosed three vulnerabilities in Paperclip, an open-source AI agent orchestration control plane, that exposed sensitive data and allowed unauthenticated command execution on servers and developer machines. Two issues were rated critical and one carried a CVSS score of 10.0. The flaws include self-registration and CLI authorization weaknesses, missing access checks, and a DNS rebinding risk in local development mode, all of which have been patched.

🔒 Vendors HashiCorp, Veeam, and the Django Software Foundation have released fixes for 11 vulnerabilities affecting Terraform MCP Server, Veeam Service Provider Console, and Django. The most severe include an unauthenticated credential-exposure bug in Veeam (9.5), a cross-tenant token-reuse issue in Terraform MCP (10.0), and a GeoDjango spatial lookup flaw that can write files or trigger code execution. Operators are advised to upgrade to Terraform MCP Server 1.1.0+, Veeam 9.3.0.35057, and Django 6.0.8 / 5.2.17; exposure depends on configuration and none of the flaws show public exploitation as of August 5, 2026.

🔍 OpenSourceMalware has identified a refined EtherHiding-style dead drop, dubbed NullReceiver, embedded in two trojanized npm packages, bianira-ui and fluid-type-ui. The technique encodes a C2 IP address directly in the recipient address bytes of an otherwise empty Ethereum transfer, allowing malware to decode the C2 from an attacker's wallet outbound transaction. The packages were published on July 28, 2026 and have been downloaded a few hundred times before removal from npm.

🛡️ This article argues that existing Zero Trust controls designed for human principals fail when applied to software agents. It proposes the Agent Access Model (AAM), which enforces short-lived, task-scoped, sender-constrained credentials, inline enforcement in the harness and network, and a Trust Ratchet that only reduces capabilities. The piece outlines architecture components—Agent Identity Broker, Task-Scoped Access Engine, Mediation Layer—and operational loops for logging and grant review to keep agent authority tightly bounded.

🚀 Filestore, Google Cloud’s first-party NFS file service, now runs on Colossus, Google’s distributed storage system, to deliver greater scalability, flexibility, and operational efficiency. The update decouples capacity from performance so IOPS can be provisioned independently, and offers deep GKE integration including CSI driver support and multishares for smaller persistent volumes. Backed by Colossus, Filestore targets high-concurrency AI and agentic workflows with NFS-based shared workspaces, improved failure recovery, and integrated security via IAM, UIDs/GIDs, and IP ACLs.