On Call Brief – Week of August 16–22, 2026

This week's top stories

1. Anthropic confirms Claude is down in major outage affecting multiple services

• Category: Deep Dive
• What happened: Claude is currently facing a significant outage, leading to login issues and degraded performance for users across multiple Anthropic services.
• Takeaway: This outage may affect any services relying on Claude, potentially disrupting operations and user access.
• Source: Bleeping Computer
• Tags: anthropic claude

2. New AmnesiaStealer macOS malware hijacks browser sessions via remote control

• Category: Deep Dive
• What happened: AmnesiaStealer is a new malware targeting macOS users that employs ClickFix attacks to hijack browser sessions. It features a streaming module enabling attackers to interactively control the victim's web browser, posing a significant security risk.
• Takeaway: This malware could lead to unauthorized access to sensitive information and accounts, making it crucial for operators to ensure security measures are in place for macOS environments.
• Source: Bleeping Computer
• Tags: amnesiastealer macos malware

3. Cloudflare: 11 scheduled maintenance windows (Hong Kong, Ashburn, Atlanta, Amsterdam (+7 more))

• Category: Community
• What happened: Cloudflare is conducting scheduled maintenance across multiple datacenters between August 17-20, 2026, affecting TXL (Berlin, 00:00-04:00 UTC Aug 17), ORD (Chicago, 08:00-11:00 UTC Aug 17), YUL (Montréal, 08:00-12:00 UTC Aug 17), SJC (San Jose, 08:00-15:00 UTC Aug 17), HKG (Hong Kong, 17:00 Aug 17 - 00:00 Aug 18 and again 17:00 Aug 19 - 12:00 Aug 20), DAC (Dhaka, 17:00-22:00 UTC Aug 17), AMS (Amsterdam, 00:00-06:00 UTC Aug 18), MAD (Madrid, 01:00-04:00 UTC Aug 18), ATL (Atlanta, 06:00-10:00 UTC Aug 18), and IAD (Ashburn, 10:00 Aug 18 - 06:00 Aug 19). All maintenance windows will result in traffic rerouting that may increase regional latency for end-users. Operators should monitor latency metrics during these windows, and customers using PNI (Private Network Interconnect) or CNI (Cloudflare Network Interconnect) connections should prepare for potential brief interruptions to direct connections at these facilities.
• Worth reading: Operators should anticipate increased latency and potential traffic failover during the maintenance window, which could affect service availability for users in the Ashburn region.
• Sources: Cloudflare Status
• Tags: hkg iad ashburn atlanta ams dac dhaka san-jose status-page-maintenance

4. Guardrail and jailbreak testing for customer-facing chatbots

• Category: Deep Dive
• What happened: The article discusses the importance of guardrail and jailbreak testing for customer-facing chatbots to prevent them from exceeding their intended scope, revealing sensitive information, or executing malicious instructions. It highlights the potential security and governance issues that can arise if chatbots are not properly tested and monitored.
• Takeaway: Chatbots can pose security risks if not properly managed, potentially leading to data leaks or malicious exploitation. Implementing guardrail and jailbreak testing is crucial for maintaining security standards in customer interactions.
• Source: Security Boulevard
• Tags: chatbots

5. Flock ALPR Expansion Raises Privacy Concerns in Traffic Enforcement

• Category: Deep Dive
• What happened: The article discusses the use of Flock cameras and automated license plate readers (ALPRs) in public safety and traffic enforcement. It highlights concerns over the expansion of surveillance from finding stolen vehicles to everyday traffic enforcement, emphasizing issues like data retention, sharing, and the risks associated with device tracking. The discussion includes implications for privacy and the potential for mission creep in the use of surveillance technology.
• Takeaway: Operators should be aware of the implications of using surveillance technologies like Flock cameras and ALPRs, particularly regarding data privacy and the potential for misuse in everyday enforcement scenarios. This could affect compliance and operational policies.
• Source: Security Boulevard
• Tags: flock-cameras alpr surveillance traffic-enforcement

6. Kubernetes NetworkPolicy Checklist: Locking Down Namespace Isolation

• Category: Community
• What happened: The article discusses the importance of implementing Kubernetes NetworkPolicies for namespace isolation to prevent lateral movement between pods. It highlights a common oversight where teams fail to apply a default-deny-all policy, leaving all pods accessible to each other. The author provides a checklist for establishing proper isolation, including steps for default-deny ingress and egress, allowing DNS egress, and managing intra-namespace traffic. The checklist aims to ensure that Kubernetes networking is secure and compliant with best practices.
• Worth reading: Failure to implement proper NetworkPolicies can lead to security vulnerabilities, allowing compromised pods to access sensitive services. This checklist serves as a guide to mitigate such risks and ensure compliance with security standards.
• Source: dev.to (Kubernetes tag)
• Tags: kubernetes networkpolicy calico

7. DevOps'ish 322: Linux wireless shuts the door on AI slop patches, KYAML fixes the Norway Bug, and more

• Category: Community
• What happened: KYAML, a strict subset of YAML designed to eliminate parsing ambiguities like the Norway Bug (where "NO" is interpreted as a boolean false instead of the country code for Norway), has entered beta and will become the default parser in Kubernetes 1.35. The specification enforces stricter syntax rules including mandatory double-quoting of strings and required use of braces and brackets for data structures to ensure consistent parsing behavior across different YAML implementations. Operators should begin testing their Kubernetes manifests against KYAML parsing rules before the 1.35 release to identify any configurations that rely on loose YAML interpretation. In related news, Linux Wireless maintainer Johannes Berg announced the project will disregard AI-generated patches due to quality concerns. Source attribution is from DevOps'ish newsletter issue 322.
• Worth reading: The introduction of KYAML could affect how Kubernetes YAML files are written and parsed, potentially improving reliability. The stance against AI-generated patches in Linux Wireless highlights ongoing debates about code quality and maintenance, which could influence how teams approach code contributions and reviews. Understanding these shifts is crucial for maintaining high standards in production environments.
• Sources: DevOps'ish
• Tags: kyaml linux-wireless

8. Agent Shell working_directory ignored; footer reports requested cwd (workspace is a git subdirectory)

• Category: Community
• What happened: The Agent Shell tool in Cursor IDE fails to apply the specified working_directory when the workspace is a subdirectory of a larger git repository. As a result, commands executed in the shell do not reflect the intended working directory, leading to errors when using git commands that rely on the correct path. The issue persists in version 3.16.17, and users are advised to use inline cd commands as a workaround.
• Worth reading: This bug can lead to confusion and errors when executing git commands from within the Cursor IDE, potentially affecting development workflows that rely on accurate directory paths.
• Source: Cursor Forum
• Tags: cursor git agent-shell

9. CMMC 2.0 Audits: Boost Defense Compliance with Lazarus Alliance

• Category: Deep Dive
• What happened: The CMMC 2.0 framework is set to enhance compliance for defense contractors by focusing on streamlined assessments and rigorous protections for controlled unclassified information (CUI). The rollout emphasizes improved Level 2 validations over Level 3 mandates, which will affect most suppliers. Lazarus Alliance offers guidance for organizations to adapt to these changes.
• Takeaway: Defense contractors must prepare for increased compliance requirements under CMMC 2.0, which could impact their operational processes and security measures.
• Source: Security Boulevard
• Tags: cmmc-2.0 defense-compliance lazarus-alliance

10. Stop manually babysitting your MCP deployments

• Category: Community
• What happened: The article discusses the challenges of managing Model Context Protocol (MCP) servers and introduces Alpic, a tool designed to automate and manage the lifecycle of MCP deployments. It emphasizes the importance of treating MCP servers as managed infrastructure rather than scripts, highlighting features like programmatic lifecycle management, telemetry feedback loops, and secure environment handling. Alpic aims to reduce the operational burden on developers by enabling agents to autonomously manage deployments and monitor performance.
• Worth reading: Adopting Alpic could significantly streamline MCP server management, reducing manual overhead and improving deployment reliability. This tool may help teams avoid common pitfalls related to environment parity and secret management, ultimately leading to more stable and secure deployments.
• Source: dev.to (DevOps tag)
• Tags: mcp alpic ai-agent automation

CVE & Security

1. CVE-2026-59124 Microsoft High Performance Computing (HPC) Pack Remote Code Execution Vulnerability

• Category: Security / Patch
• What happened: Microsoft has released a security update addressing a remote code execution vulnerability in the High Performance Computing (HPC) Pack. Users are advised to install the update promptly to mitigate potential risks.
• Do this Monday: This vulnerability could allow attackers to execute arbitrary code on affected systems, posing a significant security risk. Immediate patching is recommended to protect production environments.
• Source: Microsoft MSRC Security Update Guide
• Tags: cve-2026-59124

2. SafePal Data Breach: 39,798 Customer Records Stolen and Listed for Sale

• Category: Security / Patch
• What happened: SafePal has reported a data breach that has affected approximately 39,798 customers. A vulnerability was exploited to access customer order information, which is now being sold by a threat actor.
• Do this Monday: This breach may lead to increased risk of phishing attacks and identity theft for affected customers. Operators should be aware of potential fallout and customer trust issues.
• Source: Bleeping Computer
• Tags: safepal data-breach

3. RHSA-2026:55442: Important: bind9.18 security update

• Category: Security / Patch
• What happened: An important security update for bind9.18 is now available for Red Hat Enterprise Linux 9. This update has been rated as having a significant security impact, with detailed severity ratings provided through the CVSS base score linked to the associated CVEs.
• Do this Monday: This update addresses vulnerabilities in bind9.18 that could affect the security posture of systems running Red Hat Enterprise Linux 9 - operators should prioritize applying this update to mitigate potential risks.
• Source: Red Hat Security Advisories (RHSA)
• Tags: bind9 cve-2026-55442 red-hat

4. RHSA-2026:55432: Important: curl security update

• Category: Security / Patch
• What happened: Red Hat has released important security updates for curl across Red Hat Enterprise Linux 9 and 10, addressing vulnerabilities with significant security impact. The advisories RHSA-2026:55432 and RHSA-2026:55450 apply to RHEL 10, while RHSA-2026:55439 targets RHEL 9. Operators running affected RHEL versions should apply these curl security updates promptly, consulting the CVE links in each advisory for detailed CVSS severity ratings and specific vulnerability information. The multiple advisories for RHEL 10 may indicate different minor versions or architectures requiring separate update packages.
• Do this Monday: This update may require immediate attention to ensure that curl installations on Red Hat Enterprise Linux 10 are patched against known vulnerabilities, which could affect system security.
• Sources: Red Hat Security Advisories (RHSA)
• Tags: curl cve-2026-55432 red-hat cve-2026-55439 cve-2026-55450

5. Wireshark 4.6.8 Released, (Sun, Aug 16th)

• Category: Security / Patch
• What happened: Wireshark version 4.6.8 has been released, addressing 28 vulnerabilities and 25 bugs. This update is crucial for maintaining security and stability in network analysis.
• Do this Monday: The release includes critical security fixes that may affect network monitoring and analysis tools in production environments - operators should update to mitigate vulnerabilities.
• Source: SANS ISC
• Tags: wireshark

6. RHSA-2026:55447: Important: libXfont2 security update

• Category: Security / Patch
• What happened: Red Hat has released important security updates for libXfont2 affecting Red Hat Enterprise Linux 9 (RHSA-2026:55447) and Red Hat Enterprise Linux 10 (RHSA-2026:55448), addressing vulnerabilities rated as having significant security impact. While the advisories do not specify CVE numbers in the provided information, operators should prioritize patching libXfont2 packages on affected RHEL 9 and RHEL 10 systems, particularly those running X11 display servers where libXfont2 is used for font rendering. SRE teams should review the detailed severity ratings available through the CVE links in each advisory and plan update windows accordingly, noting that the "important" rating typically indicates exploitable vulnerabilities that could lead to system compromise. Both advisories are from Red Hat Security Advisories and should be treated with equal priority across both major RHEL versions.
• Do this Monday: This update may require immediate attention to mitigate potential security risks in production environments running Red Hat Enterprise Linux 9 - ensure systems are updated promptly.
• Sources: Red Hat Security Advisories (RHSA)
• Tags: libxfont2 cve-2026-55447 red-hat cve-2026-55448 red-hat-enterprise-linux

Releases

1. Microsoft blames AI for delayed Exchange update, can’t say when it will arrive

• Category: Release
• What happened: Microsoft has announced a delay in the release of the Cumulative Update 1 (CU1) for Exchange Server Subscription Edition due to additional work caused by AI tools used for bug-finding. The company has not provided a specific release date, citing the need to prioritize security and ensure that the update is stable. This delay raises concerns for Exchange administrators who rely on timely updates to maintain their systems.
• Do this Monday: The delay in CU1 for Exchange SE could affect organizations relying on timely updates for security and functionality. Administrators may face challenges in managing updates if they are not released as expected, potentially leading to increased workload and security risks.
• Source: The Register (Software)
• Tags: exchange-server microsoft cumulative-update

2. AWS Open-Sources Dogwood, Extending Cedar to Govern Sequences of Agent Tool Calls

• Category: Release
• What happened: AWS has released Dogwood, an open-source policy language that enhances Cedar by incorporating temporal conditions. This allows rules to consider an agent's previous tool calls rather than just a single request. Dogwood supports features like approvals, rate limits, and running totals, and is available under the Apache 2.0 license. However, the reference interpreter is not yet ready for production use.
• Do this Monday: The introduction of Dogwood could impact how policies are managed for agent tool calls, potentially improving governance and control over sequences of actions. However, the lack of a production-ready interpreter may limit immediate adoption.
• Source: InfoQ DevOps
• Tags: aws-dogwood cedar agent-core

3. GitHub Copilot’s Latest Update Bets on Model Choice, Not Model Loyalty

• Category: Release
• What happened: GitHub Copilot's latest update introduces two new AI models, Kimi K3 and MAI-Code-1.1-Flash, enabling developers to switch between models on a per-turn basis within their IDE to select the best option for specific tasks rather than committing to a single model. Separately, operators running large language models via Ollama should be aware that advertised maximum context lengths (such as 262k tokens) may not reflect actual runtime limits enforced by the Ollama environment, which can cause failures at significantly lower token counts like 33k. DevOps teams using Ollama for LLM operations should verify and configure runtime context limits explicitly rather than relying on model specifications to avoid unexpected context window failures during production use. These updates come from DevOps.com and dev.to respectively.
• Do this Monday: The introduction of multiple AI models and the ability to switch between them on-the-fly may lead to challenges in tracking which model generated specific code segments, potentially complicating debugging and code reviews. The consolidation of plugins could streamline development workflows, but teams need to be aware of the implications of model choice on code maintainability.
• Sources: DevOps.com, dev.to (DevOps tag)
• Tags: github-copilot openclaw

4. Fluent Bit 5.1.1

• Category: Release
• What happened: Fluent Bit version 5.1.1 includes several updates and fixes, such as improvements to package index tree URLs, resolution of persistent chunk issues during hot reloading, and various workflow updates. Notable fixes address buffer size underflow on startup for the S3 output and stale task dispatching in the engine. The release also includes updates to several GitHub actions used in the project's workflows.
• Do this Monday: This release may affect users relying on the S3 output and those experiencing issues with task dispatching. It's advisable to review the changes for potential impacts on existing configurations and workflows.
• Source: Fluent Bit releases
• Tags: fluent-bit

5. Google SecOps SOAR Version 6.3.98 Released with Bug Fixes

• Category: Release
• What happened: Google SecOps SOAR version 6.3.98 is being rolled out, including internal and customer bug fixes. The rollout is occurring in phases across specified regions.
• Do this Monday: This release may improve stability and performance for users of Google SecOps SOAR, but the impact of the bug fixes is not detailed.
• Source: Google Cloud Release Notes
• Tags: google-secops-soar

6. Model Catalog

• Category: Release
• What happened: The Model Catalog feature in MLflow allows for per-provider model catalog files that are updated weekly through continuous integration.
• Do this Monday: This update could streamline model management and deployment processes for teams using MLflow, enhancing operational efficiency.
• Source: MLflow releases
• Tags: mlflow model-catalog

Lightning links

• Large-scale DDoS attacks disrupted Threema secure messaging service (Bleeping Computer) -- Threema faced severe disruptions from multiple large-scale DDoS attacks, impacting service availability.
• RHSA-2026:55426: Important: 389-ds-base security update (Red Hat Security Advisories (RHSA)) -- A critical security update for 389-ds-base is now available, addressing significant vulnerabilities.
• RHSA-2026:55431: Important: vim security update (Red Hat Security Advisories (RHSA)) -- An important security update for vim is available, rated with significant security impact.
• How PGSimCity Turns PostgreSQL Complexity into a Virtual City 3D Simulation (InfoQ DevOps) -- PGSimCity is an open-source tool that visualizes PostgreSQL mechanics as a 3D simulation.
• Stop Leaving Throughput on the Table: Practical TCP BBR Congestion Control on Linux (dev.to (DevOps tag)) -- Learn how to implement BBR for better TCP congestion control on Linux, enhancing throughput.
• Own Your Workflow: An Open Source Automated Proxy Generation and Cold Storage Tiering MAM (dev.to (DevOps tag)) -- HuskHoard is an open-source tool that automates media workflows with proxy generation and storage tiering.
• Implementing Idempotent Retry Strategy for Failed Notification Systems (dev.to (SRE tag)) -- Explore strategies for retrying failed notifications using durable idempotency and queue management.
• “Continue in fresh chat” - spawn a linked agent with a clean context window (Cursor Forum) -- A proposed feature for Cursor IDE to improve context management during lengthy discussions.

Human Stories

When Claude went down this week, it reminded me how quickly we forget that every abstraction we build on is just someone else's infrastructure - and someone else's incident. The pattern across these stories isn't just about technical controls, though we see plenty of those: the Kubernetes network policies we skip implementing, the YAML parsing edge cases that turn country codes into booleans, the guardrails we need for customer-facing LLMs. It's really about the gap between what we think our systems do and what they actually allow. AmnesiaStealer doesn't exploit some zero-day vulnerability - it just tricks users into granting the access our browsers are designed to give. Flock's ALPR expansion shows how surveillance infrastructure, once deployed for one purpose, inevitably expands in scope because nothing in the technical architecture prevents it. We spend so much time hardening our runtime environments that we sometimes miss the simpler truth: most incidents happen in the space between intended behavior and actual behavior, and the only way to close that gap is to test the boundaries ourselves before someone else does.

Also worth reading

Agent Shell working_directory ignored; footer reports requested cwd (workspace is a git subdirectory) (Cursor Forum)

The Agent Shell tool in Cursor IDE fails to apply the specified working_directory when the workspace is a subdirectory of a larger git repository. As a result, commands executed in the shell do not reflect the intended working directory, leading to errors when using git commands that rely on the cor