Week 47 · Nov 16–22, 2026, snapshot Nov 16, 2026. Apple-platform stories are highlighted in orange.
Telegram launched widgets on November 10, letting any message become an app — a day later researchers at V12 demoed remote code execution, first on Android and then on the supposedly patched Windows client, blaming an unsandboxed JavaScript engine running inside messages. Kaspersky warned on November 13 that it suspects the flaw is already being exploited to deliver WebRAT malware through malicious channels, and advises users not to open widgets until a fix ships. Telegram's CEO disputes that receiving a message alone can hack anyone, saying an attack needs the user to start a third-party service. SecurityWeek reporting.
Dutch consultancy Midnight Blue found six flaws in Danfoss AK-SM800A refrigeration controllers — five in web handlers that need no login and run as root, plus one in the gateway that enforces the Strict mode login protection. No patch exists, the newest firmware 5.4.0 does not fix them, and no CVE identifiers have been assigned yet. Belgium, Ireland and Germany told operators to take the controllers' management interfaces off the internet, since the same controllers were targeted in the October attacks on French supermarkets. The Hacker News reporting.
Amazon's threat intelligence team captured AI agents — likely running on the open-source ARTEX pentesting framework — probing internet-exposed AI apps such as LiteLLM, One API and Dify to steal the AI provider keys stored inside, then using the stolen keys to run the models that powered the search for more. The agents worked from a playbook of known entry points including default credentials, unfinished installs and known flaws in Langflow and RAGFlow. The FBI has flagged ARTEX abuse since at least August, and Cloudflare separately saw the tool reach a compromised CI/CD token. The Hacker News reporting.
Kaspersky's ICS CERT found the Ukrainian-origin BO Team spent months inside a Russian software developer's network after exploiting Log4Shell (CVE-2021-44228) in an unpatched VMware vCenter 7.0.3 server to gain root, then planting a hypervisor backdoor and stealing credentials across the whole infrastructure. The attackers reached the firm's GitLab, source code and code-signing server — Kaspersky believes they aimed to inject malicious code into a signed product release or steal the signing key. The firm detected the intrusion and cut them off before the attack reached its goal. The Hacker News reporting.
VulnCheck disclosed 14 flaws in Copeland's XWEB EVO refrigeration controllers, each shown to yield root on test systems — and its sensors saw exploitation attempts against one flaw less than two hours after publication, with a second flaw probed the next day. Five flaws affect the latest firmware and will never be fixed, since Copeland ended support for the XWEB EVO family in March; one lets attackers compute the daily admin password from the device's MAC address. Nearly 2,000 of the controllers are exposed on the internet, and operators are told to get them off the public internet and plan a move to supported models. SecurityWeek reporting.
Uber is notifying roughly 150,000 drivers that their personal information was stolen after hackers used compromised credentials at a third-party services provider — confirmed as AI cloud platform Fireworks AI — to access data Uber had shared with it on October 7. The stolen data includes images of driver's licenses plus names, addresses, dates of birth and government ID numbers; Uber says the provider's credentials gave no access to its other systems. Fireworks AI is the same provider whose exposed key was behind last week's Cursor marketplace breach. SecurityWeek reporting.
A new technical analysis from SecureLayer7 reconstructs how Interlock ransomware exploited CVE-2026-20131, a CVSS 10.0 deserialization flaw in Cisco Secure Firewall Management Center, for more than five weeks before Cisco shipped a fix in March — Amazon's honeypots first recorded the attacks in late January. The unauthenticated flaw lets an attacker run Java code as root through the management interface, and CISA added it to its exploited catalog in March with ransomware use marked as known. Cisco says fixed software is available for every supported FMC branch, but any system that ran vulnerable software after January 26 should be checked for compromise, not just patched. Cyber Kendra reporting.
CVE-2026-129453, a heap buffer overflow in DrayTek Vigor2960 routers running firmware 1.5.1.6, was published on November 15 with a proof-of-concept exploit already public on GitHub; it is rated 9.4 and sits alongside two sibling flaws in the same firmware, a path traversal and a command injection. DrayTek has released firmware 1.5.1.7 to fix all three, though the router line is end-of-support and the flaw is not yet in CISA's exploited catalog. Any exposed Vigor2960 management interface should be treated as high risk and upgraded or isolated. CodeBumble reporting.
A missing-authorization flaw in the AutomatorWP Pro Webhooks addon, CVE-2026-126173, rated 9.8, lets an unauthenticated attacker reset the password of any WordPress account — including administrators — when a site has a specific documented automation pairing a webhook trigger with an Update-User action. Proof-of-concept code appeared on GitHub on November 14, and Wordfence lists version 1.4.5 as the fix for all versions through 1.4.4. No exploitation in the wild has been confirmed yet, but any site with the vulnerable automation configured should update and check for account takeovers. Wordfence reporting.
Live feed holds the current week only — older stories roll off as each new week begins.