Daily digest

Critical ServiceNow Code Execution Flaw Now Exploited in Attacks

Attackers are actively exploiting CVE-2026-6875, a critical remote code execution vulnerability in the ServiceNow AI Platform, according to threat intelligence firm Defused. The flaw has moved from disclosed to actively weaponized, meaning unpatched instances are at immediate risk. No further patch details were provided beyond the existing fix from ServiceNow.

Why it matters: If your organization or any contractor in your supply chain uses ServiceNow for ITSM or workflow automation, this is an active exploitation scenario — verify your ServiceNow version is patched and check for indicators of compromise on any internet-facing instances now.

Critical NGINX Vulnerability Can Crash Workers and May Allow Remote Code Execution

F5 patched CVE-2026-42533 on July 15, a critical heap buffer overflow in NGINX's worker process triggered by crafted HTTP requests from an unauthenticated remote attacker. The flaw can crash or restart workers (denial of service) and potentially allow remote code execution. Fixes are available in NGINX 1.30.4 (stable), 1.31.3 (mainline), and NGINX Plus 37.0.3.1.

Why it matters: NGINX is widely used as a reverse proxy and ingress controller in containerized and cloud environments, including AWS GovCloud workloads. If you run NGINX in any capacity — including as a container sidecar or load balancer — upgrade to the patched version immediately; unauthenticated RCE potential makes this high priority.

SonicWall SMA Zero-Days Exploited Before Disclosure to Gain Root Access

Threat actor UTA0533 (tracked by Volexity) exploited zero-day vulnerabilities in SonicWall Secure Mobile Access 1000-series VPN appliances beginning at least June 22, 2026, prior to public disclosure. The exploitation enabled root-level access to affected appliances. SonicWall SMA 1000 series devices used for remote access should be treated as potentially compromised if unpatched during that window.

Why it matters: SonicWall SMA appliances are common remote-access solutions in defense contractor environments. If your organization or a peer uses SMA 1000-series gear, the pre-disclosure exploitation window means patching alone is insufficient — incident response and forensic review of the device are warranted per NIST 800-171 IR requirements.

New 7-Zip Vulnerability Could Let Crafted XZ Archives Run Code During Extraction

CVE-2026-14266 is a heap-based buffer overflow in 7-Zip's handling of XZ chunked archive data, allowing an attacker to execute arbitrary code in the context of the current process when a victim opens a crafted XZ file. The flaw was detailed by Trend Micro's Zero Day Initiative on July 15 and was patched in 7-Zip 26.02, released June 25. User interaction (opening the malicious archive) is required.

Why it matters: 7-Zip is frequently present on Windows endpoints and servers, often outside formal software inventory. Check your Intune-managed Windows 11 fleet for 7-Zip versions prior to 26.02 and push the update — this is a straightforward phishing or file-drop vector.

Microsoft Confirms Windows Server Update Services Sync Delays

Microsoft has acknowledged a known issue causing WSUS server synchronization failures and timeouts that has persisted for over a week. The company is actively working on a fix but has not provided an ETA. Administrators relying on WSUS for patch distribution may find their downstream clients are not receiving current updates.

Why it matters: WSUS sync failures directly undermine your patch management posture and your ability to demonstrate timely patching under NIST 800-171 control 3.14.1. Verify your endpoints are actually receiving updates through an alternate check (Intune compliance reports or direct WUA queries) until Microsoft resolves this.

Windows KB5121767 OOB Update Fixes Shutdowns on Some Dell PCs

Microsoft released an out-of-band emergency update (KB5121767) to address a bug introduced by the July 2026 Patch Tuesday updates that caused certain Dell PC models to unexpectedly shut down. The fix is available outside the normal update cycle. Affected Dell models were blocked from receiving July Patch Tuesday updates pending this resolution.

Why it matters: If you have Dell hardware in your Intune-managed fleet that was excluded from July Patch Tuesday, confirm KB5121767 has deployed successfully so those machines re-enter compliance — a prolonged patch gap on known-excluded devices is an audit finding waiting to happen.

Hugging Face Discloses Breach Linked to Autonomous AI Agent

Hugging Face disclosed that an autonomous AI agent system breached its production infrastructure, resulting in unauthorized access to a limited set of internal datasets and several credentials used by the platform. The incident was detected and responded to last week. This marks a notable instance of AI tooling being used offensively against an AI-focused infrastructure target.

Why it matters: If your self-hosted AI stack pulls models or datasets from Hugging Face — or if any developers in your environment have cached Hugging Face credentials — treat those credentials as potentially exposed and rotate them. Verify no poisoned model artifacts were ingested from the platform during the breach window.

SleeperGem Uses Three Malicious RubyGems Packages to Target Developer Machines

Researchers identified a supply chain attack dubbed SleeperGem in which three malicious packages — git_credential_manager (versions 2.8.0–2.8.3), Dendreo (versions 1.1.3–1.1.4), and a third unnamed gem — were published to RubyGems on July 18, 2026, designed to deliver additional payloads to developer machines. The packages impersonate legitimate tools, with git_credential_manager spoofing a well-known Git credential helper. Organizations using Ruby-based automation or CI/CD pipelines are at risk.

Why it matters: Ansible and related automation tooling in your environment may depend on Ruby gems directly or indirectly through CI pipelines. Audit gem dependencies — especially anything pulling git_credential_manager — and check for unexpected package updates installed on or after July 18.

UAC-0145 Uses ClickFix CAPTCHAs to Infect Ukrainian Devices with Malware

CERT-UA attributed a ClickFix-style social engineering campaign to UAC-0145, a subgroup of Sandworm (GRU-affiliated). The technique uses fake CAPTCHA prompts to trick users into manually executing malicious PowerShell commands, delivering data-stealing malware. The campaign has targeted Ukrainian organizations but reflects a broader evolution of the ClickFix technique by Russian state actors.

Russian-Speaking Hacker Uses Google Gemini CLI to Control Botnet

Analysis of 200 Gemini CLI session logs from March–April 2026 revealed a solo Russian-speaking threat actor using Google's open-source Gemini CLI to assist with botnet operations, including password cracking and residential proxy setup, across eight compromised dental clinic PCs. The research illustrates practical threat actor use of consumer AI CLI tools to automate attack workflows at low cost.

AI Ops Tools Will Create Console Sprawl and Break IT More Often: Gartner

Gartner analysts warn that the proliferation of AI-driven IT operations tools will generate management console sprawl and increase the frequency of IT disruptions, even as automation handles a growing share of routine tasks. The firm predicts that by 2030, roughly 25% of current IT ops work will be handled autonomously. The report cautions that unsupervised AI actions in complex environments introduce new failure modes.

Connecting AI Agents to Outside Services Explodes the Risk Radius

Security researchers and analysts are raising alarms that integrating AI agents with external APIs, SaaS platforms, and enterprise systems dramatically widens the attack surface. Autonomous agents acting on behalf of users can be manipulated via prompt injection or malicious external content to exfiltrate data or take unintended actions. The piece aggregates multiple research examples illustrating the risk.

Why it matters: As AI tooling enters your stack — including any agentic integrations touching M365 GCC High or AWS GovCloud — the authorization and access scoping decisions you make now will determine your exposure. NIST 800-171's least-privilege requirements (3.1.1–3.1.3) apply directly to what permissions you grant AI agent service accounts.