Blog: security, AI and technology
Important vulnerabilities and data breaches, new AI models and changes in the technology businesses rely on. What happened, who it affects and what to do about it.
Claude Haiku 5.5: prices down up to 90%, what it means for firms
Anthropic has released a new smallest Claude model and cut prices by up to 90% compared with Haiku 4.5. We look at where it pays off, how it processes data and what to watch when migrating.
Fakturownia.pl breach: what companies and their clients must do
The attack on Fakturownia affects all accounts and the customer data stored in the service. Companies, including former customers, have their own obligations to UODO and to their clients.
October 2026 Patch Tuesday: record fixes and a Windows zero-day
One flaw is already used in phishing attacks, and several critical ones affect Outlook, Exchange and domain controllers. On top of that, the Windows Server update can stall mail flow in Exchange.
OpenAI retires GPT-4o and GPT-4.1 in the API: moving to GPT-6 Luna
OpenAI has given 92 days to move from GPT-4o and GPT-4.1 to GPT-6 Luna, and tomorrow it shuts down GPT-3.5, GPT-4 Turbo and o1, among others. Check whether your integrations use these models.
Railway incident: customer environment variables may have leaked
An intruder may have read environment variables of some Railway customers, including API keys and database passwords. The company recommends that all users rotate their secrets, and the same steps apply to any PaaS platform.
PHP2Shell: two flaws in the PHP core allow server takeover without login
Two flaws in the PHP interpreter itself allow code execution on the server without logging in, and attack attempts began four days before the fixes were released. Apache with mod_php is the most exposed. We cover what to patch today, where to look for forgotten installations and what to check in the logs.
Gemini 4 Argon: availability, pricing and what it means for business
Google's new flagship model has reached the API and Gemini Enterprise. The introductory price lasts until the end of November, and on 27 October the model is switched on by default in the Gemini Enterprise app.
LiteLLM 1.107.0: malicious release stole API keys and secrets
For a few hours on 21 October, the official LiteLLM release installed a credential stealer. If your CI or containers pulled the latest version then, you need to rotate keys.
Linux ksmbd: critical in-kernel SMB server flaws, some unpatched
The SMB server built into the Linux kernel has two new, unpatched RCE flaws rated CVSS 9.0 and a series of freshly fixed ones. Only systems where ksmbd is actually running are affected.
Node.js 27 and one release a year: what about 22, 24 and 26
From Node.js 27 every version gets LTS and a new one arrives once a year. Lines 22, 24 and 26 finish their cycle under the old rules, while 18 and 20 no longer get fixes.
Meta ordered to pay $25bn: what it means for the Pixel on your site
A US court penalised Meta for the gap between its privacy promises and its practice. The same gap often appears on company websites running Meta Pixel.
Node.js: Brotli decoder RCE flaw and TLS verification bypasses
The most serious of the 10 fixed flaws, CVE-2026-111497, affects applications that decompress Brotli data from untrusted sources. Node.js 18 and 20 will not get a fix.
CVE-2026-91403: Java trusted compromised eMudhra certificates
Eight eMudhra intermediate certificates have keys that can be recovered in a fraction of a second. The JDK trusted them, so an attacker who intercepts traffic can impersonate any HTTPS server to a Java application.
GitLab CI/CD flaw CVE-2026-97229: which instances are really at risk
A crafted regular expression in a .gitlab-ci.yml file allows code execution on the GitLab server, and proof-of-concept code is public. Instances with open sign-up or public projects should act first, though every instance needs the fix.
Want this off your plate?
As part of ongoing IT care we keep systems updated and respond to new vulnerabilities before anyone exploits them.