Chinese Telecoms Retain US Presence; 4k+ OT Controllers Still Exposed Following US Water Attacks; OSAA Proposes Shared Findings AI Exchange (SAFE)

The House Select Committee on the Strategic Competition Between the United States and the Chinese Communist Party has published a report on its investigation into three Chinese telecommunications companies: China Mobile, China Unicom, and China Telecom. The US Federal Communications Commission (FCC) revoked or denied the companies' Section 214 authorizations between 2019 and 2022 due to cybersecurity concerns, years before the discovery of the Salt Typhoon cyber espionage campaign. The legislators found that "Chinese state-owned carriers remained deeply embedded in the U.S. internet ecosystem long after federal regulators had already found them vulnerable to CCP exploitation, influence, and control." The report makes several policy recommendations, including codifying the FCC’s authority to deny blanket authorizations and restrict domestic interconnection; establishing statutory authority over retained foreign adversary infrastructure; broadening Team Telecom and Information and Communications Technology Services (ICTS) jurisdiction over private commercial arrangements involving foreign controlled infrastructure; mandating entity specific FCC Covered List determinations and funding targeted rip-and-replace; "funding internet routing-security enforcement and ensuring federal agencies have the technical talent necessary to address foreign adversary threats, [and] requir[ing] interim logging, monitoring, and recording until covered infrastructure is removed or mitigated."

Analysis published by Forescout on August 5, 2026 describes the recent attacks on US water infrastructure and provides additional details and statistics about internet-exposed Rockwell Automation/Allen-Bradley programmable logic controllers (PLCs). Forescout warns that over 4,400 devices worldwide currently expose port 44818 to the internet, and that the specific MicroLogix 1400 and 1100 models being targeted use EtherNet/IP engineering protocol on that exposed port. EtherNet/IP accounts for 22% of all OT interactions on the company's honeypots, and the researchers explain that "exposing EtherNet/IP to the internet creates an unauthenticated path that, depending on device configuration, can allow attackers to obtain information about exposed assets or even write configurations on them." 70% of the exposed devices in the US are in the networks of large mobile carriers, and are likely connected via cellular modems. Of the exposed devices in cities targeted in the recent attacks, 86% were on the same mobile carrier network, connected via cellular routers, and two hosts exposed those routers' web management interfaces; one 15-device fleet of SCADA controllers "shared adjacent network addresses, similar serial numbers, largely uniform firmware, and one identical GPS coordinate pair." While certain CVEs in Rockwell/Allen-Bradley PLCs are known to be exploited, Forescout emphasizes that "the impacts described publicly could be achieved without exploiting these issues." The researchers also saw risky exposure patterns beyond PLCs, observing "expired certificates, unrenewed remote-access hostnames, wildcard coverage that hides live services, and abandoned servers [which] together indicate incomplete asset visibility on municipal utilities." Forescout recommends isolating OT assets while allowing for remote operations with Secure Remote Access (SRA), and echoes the recommendations in advisories from the FBI/EPA, the US Cybersecurity and Infrastructure Security Agency (CISA), and Rockwell. The blog post also offers a table of recommendations for threat hunting.

NVIDIA's Open Secure AI Alliance — a coalition of over 120 organizations organized to support open-source AI security, announced following the news of OpenAI's attack on Hugging Face — is requesting comments on the Shared AI Findings Exchange (SAFE). SAFE is proposed as an independent organization comprising AI developers, infrastructure providers, customers, and researchers, as well as representatives of government and other sectors, responsible for "confidentially collect[ing] and analyz[ing] AI incidents and near misses, promptly inform[ing] affected parties and turn[ing] recurring failures into shared, evidence-based controls that reduce systemic risk." The program will handle a form of coordinated vulnerability disclosure (CVD) for AI incidents, review technical issues to promote open learning, conduct incident responses proportional to risk, establish interoperability and assurance requirements for its members, and prioritize both candid reporting and public learning. The guidelines include reporting requirements, timelines for notification, mandated preservation of evidence, a framework for incident review, a model for disclosure, and an imperative to produce detailed defensive recommendations. The same day, US National Cyber Director Sean Cairncross spoke at the Black Hat conference in Las Vegas on August 4, advocating for US investment in open-source AI.

This week, the US Cybersecurity and Infrastructure Security Agency (CISA) added five CVEs to the Known Exploited Vulnerabilities (KEV) catalog, all of which have three-day mitigation due dates. CVE-2026-63077, CVSS score 9.8, is a critical deserialization of untrusted data vulnerability in on-premises versions of JetBrains TeamCity that could result in unauthenticated remote code execution via the agent polling protocol. JetBrains has released updates to address the vulnerability, so users are urged to update to JetBrains TeamCity 2026.1.3 or 2025.11.7. CVE-2026-9198, CVSS score 9.8, is a critical code injection vulnerability in IBM Langflow OSS that can be exploited to allow unauthenticated attackers to achieve full remote code execution on default Langflow deployments. The flaw affects Langflow OSS versions 1.0.0 through 1.10.0, and users are urged to update to version 1.10.1 or later. CVE-2026-34486, CVSS score 9,8, is a critical a missing encryption of sensitive data vulnerability in Apache Tomcat that allows the bypass of the EncryptInterceptor. The flaw affects Apache Tomcat versions 11.0.20, 10.1.53, and 9.0.116, and users are urged to update to Apache Tomcat versions 11.0.21, 10.1.54, 9.0.117 or later. Patches for CVE-2026-18556 and CVE-2026-18577 address a high-severity authentication bypass using an alternate path or channel vulnerability in N-able N-central. The latter CVE is the result of an incomplete patch for the former. The flaw could be exploited to gain full administrative access to an N-central console. Users are urged to update to N‑central version 2026.3.1.7.

On Wednesday, August 5, 2026, Cisco published 14 security advisories addressing 25 total CVEs: Three advisories are rated critical, five are rated high-severity, and six are rated medium-severity. The three critical advisories are the August 2026 Cisco Catalyst SD-WAN Software Security Hardening Release (five CVEs), the August 2026 Cisco IOS XE Software Security Hardening Release (seven CVEs), and an advisory for one Cisco Secure Firewall Management Center Software Authentication Bypass Vulnerability. Cisco has released updates to address these vulnerabilities. The other advisories address a denial-of-service (DoS) vulnerability in Cisco IOS XE Software SNMP; a denial-of-service vulnerability in the Blocks Extensible Exchange Protocol (BEEP) feature of Cisco IOS XE Software; a denial-of-service vulnerability in Cisco IOS Software and IOS XE Software Extensible Messaging Client Protocol; a pair of argument injection vulnerabilities in Cisco Integrated Management Controller; a static credential vulnerability in Cisco Secure Firewall Management Center Software; two denial-of-service vulnerabilities in Cisco IOS XE Software Web-Based Management Interface; a firewall rules bypass vulnerability in Cisco Terminal Services Agent; an information disclosure vulnerability in Cisco Catalyst SD-WAN Manager; an information disclosure vulnerability in Cisco RoomOS Logging Subsystem; and a cross-site scripting vulnerability in Cisco Integrated Management Controller.

Apple's Private Relay feature, which ostensibly hides an iCloud user's IP address while using a browser, fails to do so when visiting a website with certain configurations, according to researchers Tommy Mysk and Talal Haj Bakry. Private Relay does not tunnel network traffic at system level as a VPN does, but instead routes browser traffic through proxy servers at the application level. Because this feature uses the WebKit engine required in all iOS browsers, the flaw also affects Onion Browser and the researchers' own Psylo on iOS. Three features within WebKit enable this IP exposure: DNS prefetching, WebAuthn origin requests, and WebTransport connections. If a website tries to resolve a hostname for DNS prefetching, that functionality bypasses the proxy and instead uses the device's DNS path. If a website "supports, or pretends to support, passkeys," WebKit again bypasses the proxy host app and uses the operating system's credential system to issue the HTTPS request for WebAuthn, showing the user's real IP to the destination server. Connections with WebTransport use a QUIC connection "outside the web traffic that Private Relay proxies," exposing the real IP address to the server. Onion Browser at the "Silver" security level is not affected by this last leak because it disables WebTransport for WebKit under Lockdown Mode. This news follows recent coverage of a flaw in Apple's Hide My Email feature that exposed users' underlying email addresses despite Apple's awareness of the problem for over a year (described in NewsBites 28.54). Mysk cited this delay in a social media post stating that he and Bakry opted to avoid involving Apple, instead prioritizing patching Psylo and contacting the Tor Project and Onion Browser developers first, then publicly disclosing the issue.

Adform, a global demand-side platform (DSP) digital advertiser, has disclosed a security incident that resulted in the company serving ads containing malware. Upon becoming aware of suspicious activity on July 27, 2026, Adform began an investigation and activated incident response procedures, working to contain the incident, remove the malicious code, and implement additional security measures. Adform states that the malware operated only while an affected web page was open, and "was designed to interfere with certain cryptocurrency transactions involving Bitcoin, Ethereum, or Tron. It could attempt to replace a cryptocurrency wallet address displayed, entered, copied, or pasted while an affected webpage was open." The company has not observed the malware present in ads since 17:51 UTC on July 27. According to Adform's investigation, users who visited a website that was served over HTTP and used Adform technology may have also exposed their source/public IP address, visited webpage path, and visited website hostname to the threat actor. Users who visited webpages served over HTTPS were protected from this request. Adform recommends users clear their browser caches and check the details of recent transactions with the affected virtual currencies. The company has notified publishers and advertisers who use Adform technology on sites served over HTTP, and on August 5 notified the Danish Data Protection Authority (Datatilsynet). Zack Whittaker, citing Kevin Beaumont's blog on the attack published a day before Adform's disclosure, notes that using an ad blocker "can prevent pervasive tracking, surveillance, and yes, even malware, from landing on your computer." Meanwhile, a report from the Electronic Frontier Foundation (EFF) found that " Some software development kits (SDKs) provided by advertising companies to help developers monetize their apps are automatically feeding users’ location data into systems that location data brokers use to track people."

On Tuesday, August 4, Polish convenience store chain Żabka disclosed that a cyberattack compromised corporate data in an incident that occurred via a third-party contractor. Earlier this week, Żabka detected unauthorized activity on systems that are used to communicate with the franchise network. Żabka says the intruder compromised an account belonging to a third-party service provider. The company has notified Poland's Personal Data Protection Office as well as law enforcement and cybersecurity agencies. On Wednesday, August 5, Dutch luxury department store De Bijenkorf said that a cyberattack at one of its logistics partners has delayed customers' orders, returns, and refunds, and may have compromised customer data. While customers are able to place online orders, delivery may take longer than usual. An ongoing investigation is seeking to determine whether the attacker accessed customer information. These two incidents are the latest in a string of breaches targeting retailers. In July 2026, Lidl supermarkets disclosed that customer data were compromised following the breach of an IT service provider. Also in July, a cyberattack against Japan's Nichirei Logistics Group affected the food supply chain in that country, and Harrods and Louis Vuitton both disclosed cybersecurity incidents in 2025.

A data breach affecting Ohio-based revenue cycle management services provider Unlimited Technology Systems (UTS) has been confirmed to affect more than 3.8 million individuals, making it the largest healthcare-related breach reported to the US Department of Health and Human Services Office for Civil Rights (HHS OCR) this calendar year. UTS first became aware of suspicious activity on its systems in October 2025, and in July 2026 it confirmed that data had been stolen. The incident's entry in the HHS OCR database now indicates that the breach, which is classified as a hacking/IT incident of a network server, affects 3,803,750 individuals. The compromised information includes names, physical and email addresses, phone numbers, dates of birth, health insurance data, patient balance data, Social Security numbers, medical information including diagnoses, and scanned identification documents. Lifespan Physicians Group of Massachusetts, doing business as Brown Health Medical Group-MA has disclosed that the protected health information of nearly 312,000 individuals may have been compromised in a December 2025 incident. The breach has been reported to the offices of the Attorneys General for Massachusetts and Vermont. Those reports indicate that unauthorized activity was detected on a legacy file server on December 15-16, 2025. A review of that server's contents revealed that it contained names, dates of birth, contact information, Social Security numbers, driver’s license and other government-issued identification numbers, payment card numbers, financial account information, and personnel and human resources records, possibly including compensation or payroll data, licensure or credentialing information, and medical or disability-related records. The incident's HHS OCR record places the number of affected individuals at 311,760.

A Canadian citizen has pleaded guilty to charges of computer fraud, wire fraud, aggravated identity theft, and a related conspiracy for his role in the compromise of Snowflake data storage platform customer environments in 2024. Connor Riley Moucka and his alleged co-conspirators used stolen credentials to compromise accounts; they stole terabytes of information, including individuals’ non-content call and text history records, banking and other financial information, payroll records, Drug Enforcement Administration (DEA) registration numbers, driver’s license numbers, passport numbers, Social Security numbers, and other personally identifiable information. They extorted more than $2.5 million from victims and offered stolen data for sale on cybercrime forums. As part of Moucka's plea agreement, he admitted to developing software that searched compromised cloud environments for sensitive data. Moucka was arrested in Kitchener, Ontario on October 30, 2024, and extradited to the US in March 2025. In all, affected organizations reported more than $9.5 million in losses. The breaches took place between February and October 2024, and Moucka is scheduled to be sentenced in late October 2026. Maksim Silnikau, a citizen of Belarus, has been sentenced to 16 years in prison for his role as creator and leader of the Ransom Cartel ransomware-as-a-service platform. Court documents indicate that starting in 2021, the group conducted attacks against at least 18 organizations over a two-year period, following the typical ransomware script of stealing data, encrypting systems, and demanding payment in exchange for a decryption key or a promise not to leak the stolen data. Silnikau was arrested in Spain in 2024 and extradited to the US through Poland. His co-conspirators have been charged in absentia.