Category: Security & Privacy

Would you trust AI to handle your email inbox?

Posted on by Martin Brinkmann

It was inevitable. Google is rolling out new AI functionality on its Gmail service to personal Google accounts. Called AI Inbox, it is designed to “help you manage a busy inbox”, says Google.

What that means? AI is scanning emails to identify the ones that require immediate attention. The feature has its own entry point on Gmail. When you activate AI Inbox, you get two different sections:

  • Suggested To-dos: Here, the AI lists incoming emails that need your immediate attention or action. High-priority tasks are identified and the AI explains to you in bold, what you need to do.
  • Catch-up Topics: This offers summaries of “important updates across projects and topics”, especially if they are scattered in different email threads or unrelated emails.

Google is limiting the feature currently to English-language users from the United States who are subscribed to Google AI Ultra, which costs 275 Euros per month currently (three month 50 percent introductory offer may be available).

You also need to enable smart features in Gmail to make use of it. Smart Features refers to a bundle of features, including translations, Smart Compose, or personalized search.

The Pros and Cons of letting AI handle your inbox

While there are certain pros to letting AI handle your email inbox, such as saving time, prioritization, or tone and grammar help, there are significant downsides.

Besides privacy and security concerns, there is the risk of missing important emails or of costly mistakes that the AI may make when it starts to hallucinate.

Privacy aside, the best way for users who want to make use of AI to tame their inbox is to use it as a helper, not the ultimate tool on autopilot. This is true for most AI solutions and services nowadays: you always have to verify that the AI did not miss something or introduced something that should not be there or that does not exist in the first place.

Would I use AI Inbox? I would not and the reason could not be simpler: I have no desire to give AI access to my emails because of privacy. Add a medium-sized inbox to that, and I do not have a need for any AI functionality at the time of writing.

I can see AI Inbox as a useful addition in certain cases, for instance, when so many emails arrive in an inbox that humans can’t keep anymore or when someone needs AI because of a packed day and little time to manage emails.

What is your take on this? Would you use AI features on Gmail or your email service? Or do you plant to stay away from them?

What you need to know about Firefox’s new built-in VPN feature

Posted on by Martin Brinkmann

Mozilla published Firefox 149 to the stable channel this week and it comes with a bunch of new features and changes. Besides split-view, which allows users to display two webpages side-by-side in a single browser tab, Mozilla advertises a free built-in VPN as one of the main new features.

Mozilla describes the feature in the following way:

Firefox now offers a free built-in VPN. Whether you’re using public Wi-Fi while traveling, searching for sensitive health information, or shopping for something personal, this feature gives you a simple way to stay protected. Once you sign in and turn it on, you can hide your location and IP address by routing it through a secure proxy while you browse in Firefox. You will get 50 GB of protection every month, with the option to turn it on or off for specific websites. This feature is progressively rolling out in the US, UK, Germany and France starting today.

The paragraph is different when you check out the linked support page:

VPN is a built-in Firefox feature that adds privacy by routing your browser traffic through a secure proxy server and masking your IP address. The feature includes a monthly data limit of 50 GB. Firefox will notify you when you are approaching this limit with a prompt in the browser. It is available to a limited set of users during the initial rollout, starting with Firefox version 149.

The latter is accurate, as it confirms that the solution is actually a secure proxy and not a VPN. Mozilla has likely picked VPN as it is more popular. Microsoft, actually, did the same when it introduced the Secure Network feature in Edge.

The main difference between a secure proxy and a VPN solution is that the integrated proxy only protects data from a single application, in this case Firefox.

Once activated, Firefox will route all traffic through the proxy. This protects the device IP of the user and improves privacy and security.

Mozilla says that Firefox users get 50 gigabytes of free traffic per month. This is ten times the amount that Microsoft gives Edge Secure Network users for free each month.

Another difference between the two solutions is that Mozilla relies on its own partner network for the feature, whereas Microsoft partnered up with Cloudflare.

Mozilla says that it does not log visited websites or “the content of your communications”. It does “collect technical data”, which it says is “needed to provide, maintan, and ensure the performance and stability of the service”. It also collects interaction data to “understand usage of the feature and help guide improvements”.

The feature is rolling out to users in the US, UK, Germany, and France only at the moment. You see a VPN icon in the address bar once it is available. A click displays the option to start using it.

Note: You do need to sign in to a Mozilla account to use the proxy. Once that is out of the way, you can complete the onboarding process. Users who do not want to use it can right-click on the icon to remove it from the toolbar.

Toggle browser.ipprotection.enabled to TRUE on about:config to enable it immediately, or set it to FALSE to disable the feature.

Now You: do you use a proxy or VPN when you are on the Internet?

Google Chrome 146: Security update fixes two vulnerabilities that are already exploited

Posted on by Martin Brinkmann

It is this time of the week again. Google has just released a security update for its Chrome web browser to patch two security issues with known attacks in the wild.

The update, which is available for Chrome on all desktop platforms and for Android, addresses two security issues. Google rates both with a severity rating of high.

The first issue is an out of bounds write in Skia, the specialized 2D graphics engine that is responsible for nearly everything that you see on the screen. It draws shapes, renders text, or displays images.

The second vulnerability is an inappropriate implementation in V8, another core component of all Chromium-based browsers. It is Google’s open source JavaScript and WebAssembly engine.

Google writes:

[N/A][491421267] High CVE-2026-3909: Out of bounds write in Skia. Reported by Google on 2026-03-10

[N/A][491410818] High CVE-2026-3910: Inappropriate implementation in V8. Reported by Google on 2026-03-10

Google is aware that exploits for both CVE-2026-3909 & CVE-2026-3910 exist in the wild.

Most unmanaged Chrome installations should receive the update automatically. You can speed it up by loading chrome://settings/help, if Chrome is open. Windows users may also run winget upgrade google.chrome.exe from the command line to upgrade the browser without opening it.

Expect upgrades for other Chromium-based browsers in the coming hours and days as well, as all use the very same components.

KeePass 2.61 is out: here is what is new

Posted on by Martin Brinkmann

You probably know that KeePass is still my favorite password manager and that I do not save passwords in a browser or cloud-based location. It is a free Windows-based local password manager that does not restrict passwords and can be extend easily thanks to its open system. Other developers have created apps for all kinds of operating systems.

KeePass 2.61 is the latest version that got released earlier today. The new version adds new features and improvements, including several that make the password manager more versatile or secure.

As always, while you can configure KeePass to inform you about updates, you do need to download the new version from the developer website manually, as it does not include automatic update functionality. The new version should upgrade without any issues.

The main improvements of KeePass 2.61

One of the main improvements is update-related. Checks for new updates are now performed before a database is opened. Furthermore, if the master key prompt is opened, it will now also indicate that an update is available with an icon. You can toggle the feature under Options > Advanced.

The built-in one-time password generation capabilities have received several changes:

  • White-space characters are now automatically removed when pasting shared secrets, if the encoding is Base16/Hex, Base32 or Base64.
  • New buttons in the one-time password generator to copy the passwords to the clipboard.
  • The settings dialog supports displaying history entries now.

Other than that, you get improved saving of active databases to local files, multi-location/file synchronization options, and multiple attempts at entering the master key when a database is exported. Previously, users had to re-open the option to try again if the master password was incorrect.

The changelog lists a solid number of improvements next to that, which are mostly minor changes. One of the main changes is that searches are now more tolerant by default in almost any location. You can check the full list on the linked at the top.

Ultimately, KeePass 2.61 doesn’t try to fix what isn’t broken; instead, it polishes the edges of a tool built for those who value total sovereignty over their digital keys.

The March 2026 Android Security update is here and you should install it asap (if you can)

Posted on by Martin Brinkmann

Google released this month’s big security update for Android. It fixes a total of 129 vulnerabilities, including one that is actively exploited in the wild.

As is the case with these updates, they are not published immediately to all Android devices. Pixel devices do get them first, usually, before other manufacturers start pushing them out. Even then, your device may not receive them for weeks or even months, depending on how the manufacturer handles these updates.

Google describes the most severe of the patched issues in the following way:

The most severe of these issues is a critical security vulnerability in the System component that could lead to remote code execution with no additional execution privileges needed. User interaction is not needed for exploitation.

The vulnerability affects more than 200 different Qualcomm chips and has the identifier CVE-2026-21385.

Google does not reveal how the vulnerability is exploited in the wild, but it says that it is aware of “limited, targeted exploitation” of the issue. Users should exercise caution on devices without the March 2026 patch update.

You can check the full list of patches here. Check your manufacturer’s support website to find out when your device may be getting the update. Samsung users, for instance, find the full listing on the Samsung Mobile website.

Password

Research: It appears that AI is very bad at generating secure passwords

Posted on by Martin Brinkmann

If you can’t come up with a secure password by yourself — and don’t use a password manager for that task (which most should) — then you may have come up with the idea of asking AI to give you a hand in generating secure passwords.

Cybersecurity firm Irregular published research on how that turned out for them during tests, and the result is anything but pretty.

When it asked large language models such as Claude, Gemini or GPT to generate secure passwords, it found “predictable patterns in password characters, repeated passwords, and passwords that are much weaker than they seem”.

While individual 16 character passwords looked strong, the researchers soon discovered that generating passwords multiple times would reveal the weaknesses of the approach.

Take Claude Opus 4.6 for example. When asked to generate 50 passwords, the researchers discovered several noticeable patterns:

  • Of the 50 passwords, only 30 were unique. One password was repeated 18 times.
  • All passwords started with a latter, usually uppercase G,, almost always followed by the digit 7.
  • Character choice was very uneven, with some appearing in nearly all passwords and others rarely.
  • No repeating passwords in any of the generated passwords.

ChatGPT did not fare much better. It created passwords with strong similarities. Most passwords started with the uppercase letter V, almost half continued with an uppercase Q.

Passwords generated by Gemini showed clear patterns as well. Almost half the passwords started with uppercase K or lowercase k,, usually followed by one of the characters #,, P or 9.

All AIs tested generated predictable passwords, which make it easier for attackers to brute force them. The researchers conclude that “people and coding agents shouild not rely on LLMs to generate passwords”.

Passwords generated through direct LLM output are fundamentally weak, and this is unfixable by prompting or temperature adjustments: LLMs are optimized to produce predictable, plausible outputs, which is incompatible with secure password generation.

Conclusion

Most computer users may want to stick to password managers as the go-to apps when it comes to generating strong passwords. There are free and paid solutions, local and cloud-based, something for every use case out there.

No Login? No Problem: 5 Google Maps Alternatives That Respect Your Privacy

Posted on by Martin Brinkmann

If you have used Google Maps until now without a Google account, then you may have noticed that something is off in the past couple of days.

When I launched Google Maps today in Firefox, I immediately noticed that Google was limiting information. Listings did not include user reviews anymore among other things, and Google displayed a disheartening “You’re seeing a limited view of Google Maps” and “Get the most out of Google Maps. Sign In” message at the bottom of each listing I opened.

It appears that Google is limited access for anonymous users. While you can still look up listings, use route planning, and get ads, you won’t get what some what say is the most vital information on Google Maps: user reviews.

Read Also: Google Maps is getting a new feature that you either love or dislike

Five Google Maps Alternatives

While you could sign-in to a Google account to restore full access, some may prefer switching to a different service entirely.

Here are five good alternatives that you could try:

  • Organic Maps — Organic Maps is widely considered the “gold standard” for privacy-conscious users. It is a fork of the original Maps.me, created by the original developers who wanted to strip out all the trackers and bloatware.
  • Magic Earth — If you miss Google Maps’ real-time traffic alerts and lane guidance, Magic Earth is your best bet. It manages to offer advanced “smart” features while remaining strictly no-profile.
  • OsmAnd — OsmAnd is the most feature-dense mapping app available. It’s not just a map; it’s a professional-grade geographic tool.
  • Apple Maps — In mid-2024, Apple finally brought Apple Maps to the web (currently in beta). Unlike Google, Apple’s web version actually functions better without a login, as it currently doesn’t even support signing in to an Apple ID on the browser.
  • DuckDuckGo Maps — If you are looking for the most seamless “Google Maps-like” experience in a web browser without ever being asked to sign in, DuckDuckGo is the winner. It uses Apple Maps’ MapKit JS framework, giving you high-quality visuals without the data-tracking baggage.

There are also regional apps and maps that sometimes offer better information and services than Google Maps. Kakaomap, for example, is seen as the superior app in almost any area, if you are in Korea.

Now You: do you use a map app or service? Any app that you can recommend?

Chrome Stable Channel Update: Emergency Fix for Active CSS Exploit

Posted on by Martin Brinkmann

Google has issued an urgent security update for the Chrome desktop browser following the discovery of a high-severity vulnerability being actively exploited in the wild.

The update, which brings the Stable channel to version 145.0.7632.75 or 145.0.7632.76 for Windows and Mac, and 144.0.7559.75 for Linux, specifically addresses a “use after free” flaw within the browser’s CSS engine.

Identified as CVE-2026-2441, the bug was reported by security researcher Shaheen Fazim just days prior, prompting an accelerated rollout to protect users from potential attacks that leverage this exploit to compromise system memory.

Here are the key points from the update:

  • New Versions: The Stable channel has been updated to 145.0.7632.75/76 for Windows and macOS, and 144.0.7559.75 for Linux.
  • Zero-Day Patch: The update addresses CVE-2026-2441, a high-severity security flaw classified as a “Use after free” vulnerability in CSS.
  • Active Threat: Google has confirmed that they are aware of an exploit for this specific vulnerability existing in the wild.
  • Rapid Response: The bug was reported by researcher Shaheen Fazim on February 11, 2026, just two days before the release of this patch.
  • Rollout: The update will continue to become available to all users over the coming days and weeks.

How to install the Chrome update

Most unmanaged Chrome installations should receive the update automatically. The browser is configured to install updates automatically by default. Since this does not happen immediately, it is recommended to run a manual check for updates to speed up the process.

Open Google Chrome and select Menu > Help > About Google Chrome to do so. The browser should begin downloading and installing the security update immediately.

Windows users may also run winget upgrade google.chrome.exe to install the update from the command line without opening Chrome at all.

Note that it is highly recommended to upgrade the browser, even if it is not the main browser on the system. In short, if the browser is installed, upgrade it to protect it from potential exploits.

Adbleed: A Proof of Concept for Adblocker Fingerprinting

Posted on by Martin Brinkmann

Internet users have plenty of options to make their connections more private. Popular choices include content blocking, using VPNs, or disabling services or features that may reveal information about them.

However, in rare circumstances, it is the very tools designed to protect users that may reveal information about them.

Enter Adbleed

Adbleed is a proof-of-concept designed to highlight a specific privacy risk associated with the use of regional adblocking rules.

The tool functions by detecting which country-specific filter lists—such as EasyList Germany or Liste FR—are currently active within a user’s browser. By probing for the blocking of domains unique to these specific lists, Adbleed creates a “filter fingerprint” that can reveal a user’s likely country of origin or language preference.

This technique demonstrates that users can be partially de-anonymized based solely on their adblocking configuration, even when employing VPNs or proxies to mask their physical location.

The detection process follows three simple steps:

  • Domains: The tool uses a curated list of domains that are blocked exclusively by certain filter lists, such as EasyList Germany.
  • Probing: Adbleed attempts to load resources from these specific domains. It then looks at what is returned. Blocked requests, which happen near instantly, are what the tool is after. It measures the time it takes to get a response to distinguish blocked requests from other errors, e.g., network failures.
  • Fingerprinting: When a specific number of domains are blocked from a regional listing, Adbleed concludes that the list is active.

What does it mean? It means that a site can detect if certain regional content blocking lists are likely enabled. This adds another factor to fingerprinting attempts.

Mitigation & Protection

Here are a few suggestions to mitigate Adbleed or limit its use for fingerprinting:

  • Stick to the defaults. If you do not enable any regional lists, Adbleed won’t detect any, which in turn makes your configuration less unique.
  • Enable anti-fingerprinting: If the browser supports anti-fingerprinting techniques, make sure they are enabled.
  • Disable JavaScript or enable hard-mode blocking: This may not be practicable, especially the JavaScript part, but this should protect against this particular type of attack.
  • Use different browsers: If you use different browsers, you torpedo tracking attempts, as the trackers can’t link your activities between different apps or browsers (unless there is a common factor that is unique).

Adbleed demonstrates that the tools designed to protect users on the Internet can sometimes be used against them. It reveals how regional content blocking preferences may allow sites to fingerprint and track users. It is not an argument against content blocking, but rather a wake-up call that things are never as straightforward as they look like on first glance.

Windows updates

Six Zero-Days in the Wild: The February 2026 Windows Patch Tuesday Breakdown

Posted on by Martin Brinkmann

If January was the warm-up, February is the sprint.

Microsoft’s second Patch Tuesday of 2026 has arrived with significant urgency, addressing 59 vulnerabilities in total. While the total count is manageable, the severity is high, as it contains six zero-day vulnerabilities that are currently being exploited in the wild.

Here is the breakdown of what you need to know, what to patch first, and what might break.

The February 2026 Patch Day overview

Executive Summary

  • Release Date: February 10, 2026
  • Total Vulnerabilities: 59
  • Critical Vulnerabilities: 5
  • Zero-Days (Actively Exploited): 6 (Windows Shell, MSHTML, Word, DWM, RDP, Remote Access Connection Manager)
  • Key Action Item: Administrators must prioritize workstation patching immediately due to three “one-click” security bypasses (Shell, MSHTML, Word) that allow code execution without user confirmation. Simultaneously, restrict and patch RDP servers to prevent the active SYSTEM-level escalation exploit (CVE-2026-21533).

Important Patches

  • CVE-2026-21510 — Windows Shell Security Feature Bypass Vulnerability
  • CVE-2026-21513 — MSHTML Platform Security Feature Bypass Vulnerability
  • CVE-2026-21514 — Microsoft Office Word Security Feature Bypass Vulnerability
  • CVE-2026-21519 — Desktop Window Manager Elevation of Privilege Vulnerability
  • CVE-2026-21533 — Windows Remote Desktop Services Elevation of Privilege Vulnerability

Cumulative Updates

Product, VersionKB ArticleNotes
Windows 10, Version 22H2KB5075912ESU Only. Security updates. Fixes the VSM shutdown/restart bug introduced in January.
Windows 11, Version 23H2KB5075941Security updates.
Windows 11, Version 24H2 / 25H2KB5077181Security updates and non-security changes. Adds “Cross-Device resume” and MIDI 2.0 support.

Deep Dive: The Critical Vulnerabilities

Microsoft confirmed that six already exploited zero-day vulnerabilities are fixed after installing the cumulative updates. Attackers may exploit the issues on unpatched systems to bypass protections and gain system-level access.

Here is the critical overview:

CVE-2026-21510 (Windows Shell Security Feature Bypass)

Allows attackers to craft malicious links or shortcut files to bypass Mark of the Web (MotW) and Windows SmartScreen prompts. As a result, malicious payloads may execute on unpatched systems without the usual “Are you sure” security warnings of SmartScreen.

CVE-2026-21513 (MSHTML Platform Security Feature Bypass):

Allows attackers to bypass security prompts using malicious HTML files, if the Internet Explorer engine (MSHTML) is used for rendering. The threat is similar to the Windows Shell issue described above, as it may be used to skip security screens to run malicious code on target systems.

CVE-2026-21514 (Microsoft Word Security Feature Bypass)

The third of the feature bypasses, this exploits an issue in Object Linking & Embedding (OLE) in Microsoft Office. Attackers may use it to run malicious Word documents and sidestep certain protections designed to block the execution of risky external content.

CVE-2026-21519 (Desktop Window Manager Elevation of Privilege)

The vulnerability is a type confusion flaw in the Desktop Windows Manager (DWM). Attackers need basic access for exploitation, but if they have, they may use the flaw to elevate their privileges to SYSTEM level, which allows them to take control of the system.

CVE-2026-21533 (Windows Remote Desktop Services Elevation of Privilege)

Describes an improper privilege management flaw in Remote Desktop Protocol. Exploitation opens another route to SYSTEM privileges on unpatched system. Especially problematic in Enterprise environments, which usually use RDP a lot.

CVE-2026-21525 (Windows Remote Access Connection Manager Denial of Service)

A null pointer dereference issue in the VPN / Dial-up manager. A local attacker, even with low privileges, may use the issue to crash the service repeatedly.

Significant Changes in the February 2026 updates

  • The Virtual Secure Mode (VSM) restart loop bug is fixed.
  • Cross-Device resume arrives in Windows 11. When a phone is paired with the Windows system, its recent activities are now displayed in Start. You can continue those. Requires the latest Link to Windows app.
  • Native MIDI 2.0 support. The new protocol is now supported, which creators and audio engineers may take advantage of.
  • The Secure Boot change is entering the targeting phase. In this phase, Windows can determine whether the device’s UEFI is compatible with the upcoming certificate rotation. If it is, it will be queued to receive the actual update in the coming months. No user action required.

First Steps: Your Patch Tuesday Strategy

  1. Patch the six zero-day vulnerabilities immediately. Start with user workstations.
  2. If you paused updates in January because of the VSM restart loop bug, deploy this month’s cumulative update to get it fixed.