Security news weekly round-up - 25th September 2026

By the looks of it, AI is not going anywhere anytime soon. The same applies to cyber criminals; some people are just bad and love to wreak havoc due to means best known to them. While solutions are being created to ensure our online accounts are safe from compromise, some can still find ways to subvert them.
‎
‎In the midst of all these, there are fears that AI could cause a doomsday, a ClickFix attack is still a threat, Android malware is still a thing, and you should be careful of the links that you click online.

Google’s Gemini is the latest AI model to hack other companies

There are no surprises in this one. Even the author, from the choice of words in the article's title, did not find the situation surprising. Still, you need to know about it.

Here is a piece of what happened:

Similar to OpenAI’s breach of Hugging Face, the Gemini hacks were less noteworthy for being particularly sophisticated and more for the fact that they were conducted by an AI model.

These breaches took place during cybersecurity testing by a company called Irregular. In one case, Gemini simply guessed passwords until it gained access; in the other two, it found credentials in a public repository.

An undercover Google analyst infiltrated a notorious supply-chain hacking gang

From the things that you can see in a movie to reading about it in the real world. How fun is that? Wow!

Here is a quick one from the article:

Google’s undercover analyst, Larsen says, gained access to a server where TeamPCP was storing its trove of credentials stolen from its many victims: the usernames, passwords, and access tokens it had obtained through its hacking and seemingly planned to use to extort target companies. So Google’s team decided to take action to warn victims and prevent TeamPCP’s ransom scheme.

Muse, Meta’s extraordinarily privileged AI assistant, has a serious 0-day

The introduction of the article noted that Muse was built with privacy and security in mind. This attack changes everything.

Here is what's going on:

The zero-day allows any app or terminal command to gain access to the token that authenticates users to their Muse account. Meta developers designed the assistant so that any locally installed app or executed code, regardless of the macOS permissions it has, can change a long list of undocumented settings.

Most of them are fairly innocuous, such as controlling dark mode. One setting, however, is anything but innocuous. It allows processes to change the endpoint where transcription occurs.

Normally, it’s a server address operated by Meta. Attackers can exploit this flaw by changing the location to their own endpoint. Once that happens, the attackers have the token that gives complete control over the Muse account.

Rogue external MFA providers can steal passwords during logins

When I read this article, I was like: as a security-conscious developer, you can't consider every possible security gap during development. Even when you try your best, someone can just look at the workflow of the app, do something unexpected, and then, boom, the security of the app can come down. Then what's next? Patches or mitigations.

From the article:

Varonis found that an attacker who has already compromised a highly privileged Entra account can register a rogue External Authentication Method (EAM) as one of these external MFA providers and use it to insert a convincing Microsoft password prompt into the legitimate authentication flow.

The fake prompt captures the user's password in plaintext before the malicious provider returns a valid signed token to Entra, causing the login to complete without displaying an error.

IntelligenceA Look at AI Doomsday Scenarios That Researchers Say Could Put Humanity at Risk

All the listed doomsday scenarios are something that we all should be concerned about. Although, you may argue that some are overstretched. I will say: it's better to be proactive than reactive.

The following is an example:

Some AI researchers fear a human extinction event could result from “misalignment” — that is, when AI models act without human authorization, coordinate with other models or evade oversight. It would develop not from malicious intent — AI is not human, nor is it good or evil, after all — but simply because humans might get in the way of the model accomplishing its goal.

New Android malware RemControl steals banking PINs and blocks removal attempts

I feel sick when I read these types of articles. Why would I work for my money and someone will sit down, design an app that's aimed at stealing that money? I still don't know why people do this!

From the article:

RemControl spreads through fake Google Play Store pages impersonating the TVTap IPTV application. TVTap is a popular third-party IPTV application that is not available on the Google Play Store, making users accustomed to seeking it from unofficial sources. This is why IPTV apps like this are among the most popular lures for Android malware distribution

Never Click This Type Of Link—It Could Expose Your Data To Hackers

Over the years, I would say that this became common knowledge to anyone interested in cybersecurity or has an awareness of some sort. Still, we all can forget things and need reminders; that's the way that I want you to view this article if you are already in the loop when it comes to this topic. Also, I have noticed some email clients have built features that warn you if you are navigating to an external link.

From the article:

Hackers and scammers have also gotten better at mimicking legitimate emails, making it possible for a fake message to look like it came from your bank, insurance company, a retailer you use, or even a friend or family member. A familiar name or logo doesn't necessarily mean the message is actually from the person or company it claims to be.

Credits

Cover photo by Debby Hudson on Unsplash.

That's it for this week, and I'll see you next time.