Hello, stay ahead with CISO Brief 🚀

🛡️ During the chaotic first 24 hours after a cyber incident, teams often communicate in ways that later become damaging evidence. Operational notes, Slack messages and emails— even if legal is copied—may not be privileged unless their predominant purpose was legal advice. Courts scrutinize whether communications were created for legal counsel or for ordinary business operations, and widespread channels or AI tools that share data externally can undermine privilege.

🔍 Recent studies show LLMs now excel at discovering zero-day vulnerabilities yet continue to produce insecure code at scale. Veracode found 44% of AI-generated code contains OWASP Top 10 issues and no model exceeded a 68% security pass rate, despite near-perfect syntax correctness. Other research from SIG, Xint.io, and 1Password’s Off-By-1 Labs similarly reports high rates of security violations and low patch success. Specialized harnesses and deterministic tooling improve outcomes, but human oversight and contextual organization controls remain essential.

🛡️ The Solicitors Regulation Authority (SRA) has issued a warning to solicitors and law firms about using AI responsibly after spotting hallucinations and data leaks. The notice emphasizes that regulated individuals remain accountable for AI outputs and must maintain appropriate human oversight, governance and secure handling of client data. The SRA highlighted risks including false case citations, potential contempt of court and breaches of client confidentiality when information is entered into public AI tools.

🛡️ Microsoft has removed the legacy Windows Management Instrumentation Command-line (WMIC) tool from Windows 11 24H2, 25H2 and recent beta builds as part of its planned deprecation. The company previously converted WMIC to a Feature on Demand and announced its eventual removal; WMI itself remains available. IT administrators are advised to migrate scripts to PowerShell, WMI COM APIs, .NET libraries or other modern tools. The change aims to reduce abuse of WMIC as a LOLBIN used by attackers for ransomware, evasion, and other malicious activities.

🛡️ The U.S. Cybersecurity and Infrastructure Security Agency (CISA) has added a critical Ray vulnerability (CVE-2025-62593) to its Known Exploited Vulnerabilities catalog, citing evidence of active exploitation. The flaw enables remote code execution via DNS rebinding attacks through browsers like Firefox and Safari and primarily affects developers running Ray in development or testing environments. Ray fixed the issue in version 2.52.0, and agencies are urged to remediate by August 20, 2026.

🛡️ OpenAI president Greg Brockman warned CISOs that organizations must adopt agentic systems to find and fix AI-related security flaws before attackers exploit them, citing lessons from the Hugging Face incident. He recommended tools like Codex and the Codex Security plugin and emphasized classic controls such as network isolation and least privilege. Analysts praised the guidance as sensible but noted it sounded self-serving and lacked discussion of liability and fail-safe measures for rogue agents. Experts called for stronger industry accountability and explicit rollback, audit, and blast-radius controls.

🔒 GitLab released out-of-cycle security updates on August 17, 2026, to fix a critical GraphQL vulnerability (CVE-2026-19478) that could let unauthenticated attackers remotely modify or delete public projects and user data. The patches apply to self-managed instances in versions 18.11.11, 19.0.8, 19.1.6, and 19.2.4; hosted GitLab.com and Dedicated are already patched. A second, High-severity issue (CVE-2026-19650) addresses a CSRF-related GraphQL multiplex handling flaw requiring user interaction.

🤖 Amazon Bedrock now supports OpenAI GPT‑5.6 models (Sol, Terra, Luna) on the bedrock-runtime endpoint and adds cross‑Region inference. The feature includes Global and Geo routing (now with US Geo support) to increase throughput and reduce inference costs. OpenAI Responses, Converse, and Chat Completions APIs are supported and integrate with Bedrock logging, CloudWatch metrics, and AWS cost reporting.

🛡️ A threat actor using the alias “TheHatman” is advertising employee databases allegedly exfiltrated from Microsoft Azure tenants of multiple large companies, claiming a total of 3.64 million records. The posted dumps, beginning July 31, target organizations such as McDonald’s, Tata Consultancy Services, Gap Inc., Vodafone, HCL, IHG, and Kyndryl and include names, emails, titles, phone numbers, addresses, and tenant account details. Several affected firms say investigations show no evidence of current breaches and that much of the data appears dated and non-sensitive, while cyber intelligence firm Hudson Rock assessed the samples as authentic and noted presence of service and admin accounts that could enable targeted attacks.

📣 Pokémon Center has notified UK and German customers that a third-party logistics provider, CEVA Logistics, suffered a cyberattack that exposed customer personal and order information. The breach affected CEVA systems between July 29 and August 1 and disrupted several European warehouses, causing shipping delays and cancellations. Pokémon Center says exposed data may include names, mailing addresses, phone numbers, email addresses, and order details, but not payment card information.

🚀 Amazon Elastic Container Registry (Amazon ECR) has increased the maximum number of replication rules per registry from 10 to 25. This change addresses limitations for complex multi-region and multi-account architectures, allowing more granular replication strategies for scenarios such as distributing images for low-latency pulls and separating production and staging targets. The increase is available in all AWS Regions where Amazon ECR is supported.

🔒 Researchers at Wiz disclosed a GitHub Actions workflow injection in Snowflake's snowflakedb/snowflake-connector-net repo that allowed attacker-controlled issue fields to be expanded into a shell run: block, exposing JIRA_BASE_URL, JIRA_USER_EMAIL, and JIRA_API_TOKEN. The vulnerability stemmed from unsafe expression expansion in .github/workflows/jira_issue.yml and incorrect event property checks, which let a crafted public issue reach the job. Snowflake patched the workflow the same day Wiz reported it and rotated the exposed Jira token; no evidence of unauthorized access or affected releases was found.

🛡️ A critical vulnerability (CVE-2026-15748) in the Forminator Forms WordPress plugin — used on 600,000+ sites — allows unauthenticated attackers to upload arbitrary files, including executable PHP, and achieve remote code execution. The flaw, present in versions up to 1.56.1, stems from improper file type validation in the handle_file_upload() function and misuse of MIME key matching combined with a public submission handler. Patch 1.56.2, released on July 31, 2026, fixes the issue; site owners should update immediately.

🔁 Amazon EC2 Auto Scaling now allows terminating up to 100 instances with a single TerminateInstanceInAutoScalingGroup API call, reducing API request volume when scaling down. Batch termination is intended for rapid scale-down scenarios like AI/ML training, container orchestrators, or event-driven fleets and validates all instances atomically before proceeding. Existing Auto Scaling behaviors such as lifecycle hooks and load balancer connection draining remain enforced for each instance. The capability is available in all AWS Regions at no additional cost.

🔧 Amazon MSK Provisioned clusters now support configuring custom domain names at the cluster level for both ZooKeeper and KRaft metadata modes. This removes the need to set domains per broker and ensures persistent endpoints across scaling, migrations, and failovers. The setting persists through cluster operations and is available for new and existing clusters in all Regions where MSK Provisioned is offered, at no extra cost.

🔍 Kaspersky researchers uncovered new components of the Cavern (CAV3RN) command-and-control framework used by Iranian-linked operators to target Israeli entities, revealing a module that switches between direct HTTPS and a Google Apps Script relay using DNS A-record responses. The modular toolkit supports extensive post-exploitation functions and minimizes forensic visibility, while additional reports show HOLLOWGRAPH abusing Microsoft 365 calendars and DNS tunneling to maintain and refresh Azure AD credentials. The findings highlight a shift to a plugin-based architecture and continued use of legitimate services to evade detection.

🚀 Amazon EC2 R8i and R8i‑flex instances are now available in the Canada West (Calgary) region. These instances use custom Intel Xeon 6 processors exclusive to AWS, offering up to 15% better price-performance and 2.5x memory bandwidth versus prior Intel-based instances. R8i provides large sizes including 96xlarge and SAP certification; R8i‑flex offers common memory‑optimized sizes from large to 16xlarge. Purchase options include Savings Plans, On‑Demand, and Spot.

🔍 The Amazon Connect Customer dashboards now include reporting on routing steps and agent proficiencies, enabling supervisors to monitor and optimize contact-to-agent matching. Managers can filter agents by assigned proficiencies, group metrics by routing steps, and track metrics such as the number of contacts queued in a routing step. These analytics are available in all AWS commercial and AWS GovCloud (US-West) regions where Amazon Connect Customer is offered.