30+ MN Community Water Utilities Impacted by Cyberattacks; IPMI Flaw Exploited to Leak BMC Password Hashes; Cisco Secure FMC Hardcoded Credential Exploited

Threat actors launched cyberattacks against operational technology (OT) systems at more than 30 municipal water systems in the US state of Minnesota. According to Minnesota IT Services (MNIT), the attacks took place on Sunday and Monday, July 26 and 27, and the affected utilities were forced to revert to manual processes. Upon learning of the cyberattacks, MNIT activated Minnesota's cybersecurity incident response capabilities, and is coordinating technical response activities across government partners; sharing threat intelligence and indicators of compromise with affected organizations; supporting incident investigation, containment, recovery, and remediation efforts; monitoring for related malicious cyber activity and coordinate with state and federal cybersecurity partners; and providing cybersecurity expertise and technical assistance to affected communities. MNIT is working with federal, state, local, Tribal, and private-sector partners to investigate the attack. Days before the attacks, the US Cybersecurity and Infrastructure Security Agency (CISA) published an updated version of an April 7, 2026 advisory regarding state-sponsored cyber threat actors targeting operation technology (OT) devices used in US critical infrastructure, including programmable logic controllers. Additionally, on July 28, 2026, CISA, along with Australia’s Signals Directorate (ASD), the UK’s National Cyber Security Centre (NCSC), and Canada’s Centre for Cyber Security, published *CI Fortify – Advice for isolating vital systems*, which "explains how organisations can isolate critical operational technology (OT) and supporting systems from other networks during cyber incidents or periods of increased cyber threat ... and provides practical guidance and strategic advice for OT owners, operators, and cyber security teams."

Researchers at Lava Security are warning that nearly 25,000 publicly exposed Baseboard Management Controllers (BMCs) leak password hashes due to a flaw that has existed since 2004 in the Intelligence Platform Management Interface (IPMI) 2.0 protocol. A BMC is typically a separate ARM processor built into the motherboard of a device such as a server, allowing administrators to remotely access and manage device health and functions even if the operating system is not working. CVE-2013-4786, CVSS score 7.5, allows a remote attacker with access to UDP port 623 to test password guesses offline, because the RMCP+ Authenticated Key-Exchange Protocol (RAKP) in IPMI 2.0 can cause the BMC to return an HMAC-SHA1 authentication code before authentication is complete. Lava explains that "weak, reused, factory-set, or predictably formatted passwords may therefore be recovered without generating the usual pattern of repeated failed logins," exposing BMCs' pivotal position outside the operating system, outside security tools' reach, and connected to management networks. During Lava's testing, more than 25% of internet-exposed IPMI hosts that returned an RAKP response before authentication also accepted an empty username with a weak password. More than 50% of BMCs in Lava's dataset were Supermicro hardware, whose unique alphabetic factory passwords could still be brute forced within hours by modern GPUs. Yakir Kadkoda, Lava's co-founder and CTO, stated to Dark Reading that Lava also found evidence that an unnamed major automotive component manufacturer had been targeted with ransomware following compromise via internet-exposed BMCs. The researchers urge administrators to cut off BMC and IPMI from the public internet and additionally block UDP port 623 at the network edge; rotate factory passwords; disable legacy and weak configurations; strictly control BMC access via private network, VPN, or bastion host; implement network access controls; and prioritize monitoring of the management network.

Cisco has published a security advisory urging users to update Cisco Secure Firewall Management Center (FMC) software to fix a static credential vulnerability that is under active exploitation. CVE-2026-20316 allows an unauthenticated remote attacker to access data available to a low-privileged user by logging in with a hardcoded credential; the flaw carries a base CVSS score of 5.3, but Cisco assesses it as high impact because it could be chained with other FMC flaws to achieve privilege elevation. FMC versions 7.0, 7.2, 7.4, 7.6, 7.7, and 10.0, regardless of device configuration, contain the static credential and must be patched. No details of the exploitation have been made public at the time of this writing, but the US Cybersecurity and Infrastructure Security Agency added the flaw to its Known Exploited Vulnerabilities catalog (CISA KEV) on July 29, with a remediation deadline of August 1. Cisco's advisory includes a command-line check for indicators of compromise (IoCs), and recommends that users rotate all credentials, keys, and certificates.

Researcher Håkon Måløy has published a blog post detailing his discovery that Microsoft Copilot is vulnerable to a prompt injection attack that results in a worm propagating through Word documents. Måløy summarizes the flaw: "Malicious instructions hidden in an externally shared document could make Copilot alter drafted or edited documents in Word and propagate the attack to new documents." A JSON-formatted malicious prompt hidden from the user as white text in a small font will still be read by Copilot when the document containing it is uploaded to Copilot for Word or discovered by Copilot in the user's OneDrive and ingested as context. The prompts frame executing the malicious instructions and copying them into a new document as relevant to a benign user task, leading to infected documents that carry trust as internally created files. Måløy disclosed this exploit to Microsoft in March 2026, and while the company successfully patched Copilot against the original prompts, the broader class of attack is still viable with adjusted prompts; after 144 days of coordination, Måløy and Microsoft agreed to a public disclosure. Måløy believes that resolving this issue "requires research rather than a single patch." He notes that "the integrity of information becomes a primary security concern in systems that integrate LLMs as part of their operating workflows," and he concludes the report by contending that cross-domain prompt injection attacks (XPIA) are difficult to mitigate because of the fundamental design of current LLMs: "The long-term challenge likely lies in designing systems in which goals and intentions also exist independently of the information being processed [...] Any system that integrates an LLM into a trusted workflow today must assume that attacker-controlled content entering the model’s context will result in compromise at some rate."

An updated version of the Model Context Protocol (MCP) was published on July 28, 2026, including security hardening and architectural changes primarily to serve the scale of enterprise use. The MCP is an open-source protocol that standardizes communication between LLMs and external tools and sources of data, initially created by Anthropic engineers in 2024 and then given to the Linux Foundation (LF) in December 2025 to be maintained under the Agentic AI Foundation (AAIF). Caitie McCaffrey at the LF states in a blog post that "this release removes the core scaling bottlenecks, finalizes the project's transition to a stateless architecture, and establishes the formal governance and security hardening required for enterprise-scale deployments." The change to stateless architecture should allow organizations to treat MCP servers like scalable HTTP services, and to run them behind standard load balancers using their existing tools. Governance changes include a formal policy and timeline for deprecating features and a new framework for extensions. The updated MCP also aligns with "how OAuth 2.0 and OpenID Connect are deployed in practice," including mandatory validation of the iss parameter to help prevent OAuth mixup attacks, and allows for central management of authorization. This is the first major update to the MCP since it was transferred to the AAIF.

On Monday, July 27, Apple released updates to address vulnerabilities across its product line. In all, the updates address 187 vulnerabilities: 87 flaws were fixed in iOS 26.6 and iPadOS 26.6; 155 flaws were fixed in macOS Tahoe 26.6; 138 flaws were fixed in macOS Sequoia 15.7.8; 127 flaws were fixed in macOS Sonoma 14.8.8; roughly 100 flaws were patched in each tvOS 26,6, watchOS 26.6, and visionOS 26.6; and 10 flaws were addressed in Safari 26.6. In an Internet Storm Center diary, Dr. Johannes Ullrich writes, "In June, Apple announced that it may publish occasional ‘security update only’ releases. This release does not contain any significant new functionality but is also meant as a ‘prep release’ for iOS/macOS 27, as it makes some adjustments to Spotlight to get the system ready for the new major OS releases coming in the fall."

On Wednesday, July 29, Broadcom released updates to address five vulnerabilities that affect VMware ESX, VMware vCenter, VMware Workstation, VMware Fusion, VMware Cloud Foundation, VMware vSphere Foundation, VMware Telco Cloud Platform, and VMware Telco Cloud Infrastructure. CVE-2026-59309 is a critical authentication bypass vulnerability in vCenter VMware Directory Service. CVE-2026-59310 is a critical directory traversal vulnerability in vCenter Syslog server. CVE-2026-47876 is a critical out-of-bounds write vulnerability in VMware ESX VMXNET3 virtual network adapter. CVE-2026-41703 is an out-of-bounds read vulnerability in VMware ESX, Workstation, and Fusion, which is rated important for ESX and low severity for Workstation and Fusion. CVE-2026-41709 is a low severity insufficient logging vulnerability in VMware ESX. In addition to the security advisory, Broadcom has published a "Questions and Answers" document for the vulnerabilities, which supplements the advisory with “self-service information to help you and your organization decide how to respond."

On Wednesday, July 29, 2026, the US Cybersecurity and Infrastructure Security Agency (CISA) — along with other cyber authorities in the US, Australia, Canada, Czechia, France, Germany, India, Italy, Japan, South Korea, the Netherlands, New Zealand, Poland, and Slovakia — published updated guidance on creating a software bill of materials (SBOM). The *2026 Minimum Elements for a Software Bill of Materials (SBOM)* "reflect[s] current SBOM needs, while preserving the core principles of the document published in 2021 by the National Telecommunications and Information Administration (NTIA)." A draft version of the updated guidance was published in August 2025, and the version published this week incorporates responses received following a request for comment on that draft. On Thursday, July 30, CISA published open source software guidance for federal agencies. *Open Source Software: Security Principles and Practices* includes recommendations "on the use of security assessments and patching of open source software (OSS) and best practices for contributing to OSS projects. These recommendations are rooted in software development and software supply chain risk management best practices and are tailored to address the benefits and risks unique to OSS."

The Health Information Sharing and Analysis Center (Health-ISAC) has published an advisory noting that organizations within the healthcare sector are experiencing "an observed increase in successful attacks by the threat actor, ShinyHunters." The threat actor appears to follow a predictable path: using voice social engineering at the helpdesk to request assistance with an MFA reset or device re-enrollment, followed by taking control of a Microsoft Entra or Okta/Google single-sign-on (SSO) account. From there, the threat actor pivots into connected software-as-a-service (SaaS) platforms and exfiltrates data. The Health-ISAC advisory lists recommended mitigations for healthcare organizations to help protect their networks from ShinyHunters attacks at each step of the threat actor group's modus operandi. The document also provides a 30-60 day shortlist of steps to take first, which includes implementing phishing-resistant MFA for admins, helpdesk, and executives; restricting weak MFA fallbacks; implementing helpdesk reset hardening; establishing and implementing Entra/SSO conditional access baseline; implementing methods to detect SaaS exfiltration; and testing token/session revocation playbook in tabletop exercises.