Weekly from SecurityWeek Cybersecurity News Summary provides a succinct overview of key developments that may not be covered entirely in isolation but are nonetheless relevant to the broader threat landscape.
This curated summary highlights key stories about vulnerability disclosures, new attack vectors, policy updates, industry reports and other notable events to help readers stay well-informed about the evolving cybersecurity environment.
Here are this week’s highlights:
Clop’s leak page was seized by ShinyHunters in a grudge match
ShinyHunters has defaced the Cl0p ransomware gang’s Tor data leak page. The blackmail group Claims Server logs, source code, and the private keys for Clop’s Onion service were also stolen. It demanded an eight-figure payment and a public apology and threatened to expose companies that allegedly paid Cl0p as part of its Oracle E-Business Suite campaign. ShinyHunters says the attack was in retaliation for threats allegedly made by a Clop representative as part of a feud stemming from that campaign.
BragJack attack against browser AI assistants
Researchers at endpoint security company Forever made the announcement BragJacka series of bugs that allow a malicious extension to take control of the built-in AI assistants in Chrome, Edge, Opera Neon, Perplexity Comet and the Claude in Chrome extension. In any case, the wizard trusts the commands of a specific website. An installed extension could hijack this page by injecting scripts or manipulating network traffic, then sending its own prompts without user interaction. Depending on the browser, this made it possible to read emails, access local files, take screenshots or turn on the camera and microphone. The sellers paid bounties ranging from $600 to $7,000.
Worm-enabled Go implant sneaks into AI agents’ memory tools
An attacker has released malicious versions of MemTensor’s MemOS packages to npm and PyPI, including a storage plugin for the OpenClaw AI agent harness. The packages contain a previously unknown Go implant called sckit. Instead of running at installation time, the malware is launched when the Python library is imported or the NPM plugin is used. It searches for npm, PyPI, GitHub, AWS, Hugging Face and other secrets. The implant includes templates for distribution via npm, PyPI and GitHub Actions, but Semgrep says there is no evidence yet that it has spread. Aikido And StepSecurity also shared details.
AI relay networks forward Chinese traffic to Western border models
Team Cymru has found Nearly 11,000 servers running the Claude Relay Service or its successor sub2api. These open source gateways pool AI accounts so many users can share them, while model providers only see the relay and never the actual user or their location. In a US-hosted cluster, more than 4,000 IP addresses in China and Hong Kong (regions that exclude Anthropic, OpenAI, and Google) were connected to 304 relays that also reached OpenAI, Anthropic, xAI, and Google endpoints.
Infostealer protocols expose remote access keys across the US water sector
SpyCloud analyzed Stolen identity data linked to 10,000 U.S. water and wastewater utilities and the technology providers that serve them. Active infostealers were identified at 1,787 organizations and OT or remote access system credentials at 258 organizations. In one case, malware captured logins stored for approximately 167 energy metering portals on a single device at an advanced metering technology provider. Exposed credentials at the utilities themselves primarily involved remote management tools such as TeamViewer and SonicWall, as well as Fortinet management portals, although SpyCloud emphasizes that the findings reflect potential access paths and not confirmed intrusions.
CLOSEDQUORUM swaps C2 servers for commercial AI APIs
Cisco Talos documented CLOSED QUORUMa Go-based Windows implant that it believes is the first publicly documented system to leave its command and control decisions to commercial LLMs, rather than a human operator or an attacker-operated server. Up to four models (DeepSeek, Qwen, Mistral and Gemini) vote on whether to steal credentials, inject code or establish persistence. The implant then carries out the winner selection and sends LSASS dumps, browser passwords and crypto wallet data to the operator’s Discord channel. Talos has not confirmed use in the wild and the public build includes placeholder API keys, but development builds indicate that the developer is creating custom versions for individual operators.
Canonical promises workarounds for the Ubuntu kernel within 48 hours of release
Canonical is replacing Ubuntu’s separate four-week regular and two-week stable release update (SRU) cycles for the security kernel with one single two-week cycle. Since the cycles overlap, a new kernel is released every week. The company cites a sharp increase in CVE volume driven by AI-powered bug detection and the upstream kernel community becoming its own CVE numbering authority, assigning identifiers to thousands of bugs. Administrators who need fixes sooner can test release candidates from the -suggested bag before the certification exam is completed. Canonical also aims to provide workarounds or hardening instructions within 24 to 48 hours of a vulnerability being published.
Error in TDengine before authentication jeopardizes industrial telemetry uptime
Ridge Security has published Details of CVE-2026-42542a fatal flaw in TDengine, a time series database used in industrial telemetry, energy, utilities and IoT environments. An unauthenticated attacker can crash the server if a single malformed packet is sent to its RPC port. The error is an integer underflow in message parsing that occurs before authentication and results in a heap buffer overflow. The researchers only confirmed denial of service, but urge defenders to also consider underlying memory corruption. TDengine versions 3.4.0.0 to 3.4.1.5 are affected and version 3.4.1.6 fixes the problem.
The banking Trojan’s AI helper thought it was creating a quiz
Group-IB has revealed BrakeControla new Android banking Trojan offered as malware-as-a-service. It spreads through fake Google Play pages for the IPTV app TVTap, targeting customers of more than 30 banks in Western Europe, the Middle East and Canada. Once access permissions are granted, the malware displays phishing overlays over banking apps, streams the screen, logs keystrokes, and allows the operator full remote control of the device. Exposed API documentation suggests that parts of the platform were built using an AI assistant that was told it was working on a quiz and parental control app, and a phishing overlay included a full response from the AI assistant.
The Docker botnet ranks AI API keys above all other loot
ThreatDown described in detail CARBONATEa botnet that compromises Docker daemons without authentication on port 2375 and scans neighboring networks every five minutes to further spread. On each host, it installs Hermes Agent, a legitimate open-source AI agent framework, and replaces its persona file with instructions to follow operators’ Telegram commands, maintain persistence, and collect credentials, putting AI API keys first. The researchers found the operation via an exposed, unauthenticated Docker registry. Evidence of language, time zone and infrastructure supports their assessment that the operators are based in Costa Rica.
Related: In other news: Ransomware developer convicted, Plugin4Shell AI attack, critical SAP flaw
Related: In other news: InjectEave attack, SIM swapper convicted, Glasswing findings review
