News archive

[WARNING · CHECK THIS WEEK]

European Commission attacked with help of a public AI tool; check which internal apps you expose

The European Commission confirmed on 19 November 2026 a cyber campaign against some of its systems in which a publicly available AI tool was misused. According to Politico, CERT-EU, the EU institutions' security team, identified the tool as ARTEX, built to look for weaknesses in IT systems. As a precaution the Commission took several applications offline, including Sysper, the HR platform staff use for payslips, leave and timesheets.

[WARNING · CHECK THIS WEEK]

Java4Shell: OpenJDK fixes a no-login hole in Java lines 17, 21, 25 and 27; find and patch your Java

On 17 November 2026 OpenJDK published fixes for CVE-2026-114938, a HotSpot compiler flaw in four Java lines that an attacker can use over the network without credentials. Calif, the firm that found it, calls it Java4Shell. Oracle shipped its November patch update the same day, and South Korea's intelligence service NIS warned public institutions to get ready.

[WARNING · CHECK THIS WEEK]

Malware via Optimizely Web reached visitors of 178 organisations: check what scripts your site loads

On 16 November 2026 visitors of at least 178 organisations, Zoom, the BBC and HSBC among them, received malicious JavaScript through Optimizely Web, an A/B testing service, Report URI says. The code sat in each customer's own snippet, served by Optimizely; the sites needed no flaw of their own.

[WARNING · CHECK THIS WEEK]

Google Project Zero publishes bug details 90 days after report, 30 by 2028: measure your patch time

On 13 November 2026 Google Project Zero, the company's team of vulnerability researchers, switched to a disclosure policy it calls "Full 90": technical details of the bugs it reports go public 90 days after the vendor hears about them, whether or not a fix exists. The deadline drops to 60 days on 2 February 2027 and to 30 days on 3 January 2028.

[CRIT · FIX TODAY]

Synology DSM flaws let attackers read and write NAS files without an account: update now

On 10 November 2026 Synology rated advisory SA-26:19 Critical: two flaws in DiskStation Manager (DSM), the NAS operating system, work over the network without a login; one lets attackers read or write files of their choosing, the other gets past the rsync login. There is no workaround besides the update. Neither the advisory nor Tech Journal reports attacks.

[CRIT · FIX TODAY]

Apache HTTP Server: critical HTTP/2 flaw can lead to code execution – update to 2.4.71

Apache HTTP Server 2.4.71, released November 11, fixes CVE-2026-119618, rated critical by the project. A race condition in the HTTP/2 module (mod_http2) can corrupt memory and lead to remote code execution in versions 2.4.55 through 2.4.70. Cloud security firm Wiz says exploiting it needs no login, and that no exploit or attacks were known when it published.

[WARNING · CHECK THIS WEEK]

Grok Build: 35,578 AI-built apps opened for anyone – check what your staff published

Security firm Claroc reported on November 10 that of 37,571 Grok Build apps it found still online, 35,578 (94.7%) opened with no sign-in. The open apps held internal material of SpaceXSI (formerly xAI), which makes Grok, and of 30 other organisations Claroc could identify, and a hospital group's roster of more than 3,500 employees.

[CRIT · FIX TODAY]

N-able N-central: exploits for 30+ alleged flaws are public – keep on-prem servers offline

Late on Sunday, November 8, a researcher posted proof-of-concept exploits for more than 30 alleged vulnerabilities in N-able N-central, the remote monitoring and management (RMM) platform IT providers and IT teams use to run their customers' computers. On November 9 N-able declared a critical incident, shut down its hosted N-central in the Americas, Europe and Asia-Pacific, and advised on-premises customers to switch their servers off until further notice. As of November 9, no fix was out.

[WARNING · CHECK THIS WEEK]

Fake AI event invites phish Microsoft 365 accounts with device codes – block device code sign-in

Push Security found fake registration pages for OpenAI, SpaceX and Anthropic events promising ChatGPT Pro, Grok Heavy or a "SpaceX share" to anyone who signs in with a work Microsoft or Google account. BleepingComputer reported it on November 5 and TechRadar on November 6; Push Security detected and blocked visits to the pages in late October.

[WARNING · CHECK THIS WEEK]

Claude Code's "don't ask again" writes secrets into a file that gets committed – check your repos

Security firm Apiiro reported on November 7 that Claude Code stores every command a user approves with "Yes, and don't ask again" – secret included, even one taken from a config file – in .claude/settings.local.json, a file that regularly ends up in Git. Across this file, its history and other agent files, Apiiro counted 2,539 real secrets in 1,994 public GitHub repositories.

[WARNING · CHECK THIS WEEK]

Ruby gem worm takes AI coding tool credentials and plants agent files in repos – check yours

Socket analysed strutil-core 1.0.1, a RubyGems package that runs a credential-stealing worm when installed. It was public on November 1 for about five hours before RubyGems removed it. Besides cloud, SSH and registry secrets, the worm takes credentials of AI coding tools, adds its own entries to their local settings and, with stolen Git access, commits instruction files for coding agents into repositories.

[CRIT · FIX TODAY]

Zyxel GS1900 switches: an exploit used in attacks gets past the June patch – update to .3C0

Zyxel published an advisory on November 3 for an out-of-bounds write (CVE-2026-107441) in the web management of ten GS1900 switch models; fixed firmware has been available since October 22. Binarly researchers found that an exploit used by attackers still works against switches that installed Zyxel's June fix. GreyNoise reported that around August 17, attackers took data from 996 GS1900 switches in 48 countries.

[WARNING · CHECK THIS WEEK]

Hacked websites hit out-of-date iPhones and Android phones – update your staff's phones

Security firm Sekoia says APT29, a group widely linked to Russia's foreign intelligence service, planted attack code on hacked websites and lookalike domains. On iPhones and Android phones with outdated software, opening an affected page was enough: the exploit chains needed no further interaction. Sekoia, which found it in mid-October, attributes it with moderate to high confidence. Apple and Google had already patched every flaw it used.

[WARNING · CHECK THIS WEEK]

Supermarket freezers hacked again – check what you expose to the internet behind just a password

On the night of November 1–2, attackers remotely heated fridges and freezers in dozens of supermarkets in Belgium, Luxembourg and Sweden and the food spoiled, Reuters and local media report; in Belgium they switched the units to defrost mode. Stores in Ireland and Estonia report incidents too. It is the second wave after hundreds of French stores were hit on October 11–12.

[WARNING · CHECK THIS WEEK]

Stolen AI and cloud keys ran up over $100,000 in one day: find and rotate yours

Google Threat Intelligence Group and Mandiant reported on 2 November 2026 that criminals hijack companies' access to AI models ("LLMJacking"). In incidents Mandiant handled this year, fraudulent charges topped $100,000 within one day and reached $1 million at some organisations before access was cut. The usual way in was exposed or stolen cloud credentials.

[CRIT · FIX TODAY]

Drupal Commerce Core: access bypass exploited without login – update, then audit your orders

On October 28, Drupal's security team warned of active attacks on a highly critical access bypass in Commerce Core, the module that adds carts and checkout to Drupal sites. An attacker needs no account. Fixed releases are out; Drupal asks shops to go through their orders and warns the flaw may cost the seller money.

[CRIT · FIX TODAY]

Ivanti Connect Secure: three exploited flaws give root without login – patch, then check AD

CISA added two more Ivanti Connect Secure flaws to its catalog of exploited vulnerabilities on October 30. Chained with an authentication bypass listed on October 13, they give an attacker root on the VPN appliance without credentials. Volexity, which found the attacks in the first week of October, says the attackers deleted evidence from the appliance's logs and went after Active Directory secrets.

[CRIT · FIX TODAY]

C3RB3R ransomware hits Synology NAS (.L0CK3D): update DSM, close admin ports to the internet

Since early October, Synology NAS owners whose data was encrypted have been asking Chinese security vendor Qihoo 360 for help; 360 published an alert on 29 October. Synology and the researchers disagree on how attackers get in, and one victim was on the latest DSM.

[WARNING · CHECK THIS WEEK]

Binance: a hacked DevOps laptop may have started the theft. Check what one admin device can do alone

On Thursday an attacker moved crypto out of Binance hot wallets. On 30 October the exchange estimated the impact at about $6.3 billion and said its early findings pointed to a DevOps engineer's compromised laptop; its co-CEO later added that the attack was more complex and had further points of breach.

[CRIT · FIX TODAY]

PHP2Shell roots Palo Alto firewalls with exposed admin pages: restrict access, look for web shells

Dream reports that a suspected China-linked attacker chained PHP2Shell, two PHP flaws, with the Fragnesia kernel bug and got root on 24 firewalls with internet-facing management across South Korea, Taiwan, Japan and the US on October 23–24. Dream withheld the vendor; The Register, via Mezha, says the devices run Palo Alto Networks PAN-OS versions 12.1 and 12.2.