Loading video...

Video Failed to Load

Go Home

‼️🚨 This is alarming: Researchers found a one-click data exfiltration vulnerability in M365 Copilot. A single click on a trusted microsoft[.]com link let attackers pull emails, MFA codes, meeting notes, and SharePoint/OneDrive files, no permissions or second click required. Microsoft has patched it as CVE-2026-42824, rated critical.

143,911 views • 3 months ago •via X (Twitter)

35 Comments

International Cyber Digest's profile picture
International Cyber Digest3 months ago

Source:

Apropos's profile picture
Apropos3 months ago

Found another malware spreading through GitHub

Cybernorse's profile picture
Cybernorse3 months ago

Organizations must verify that CVE‑2026‑42824 is fully patched across M365 Copilot to block one click data exfiltration of emails, MFA codes, and files. This critical zero‑day reinforces the need for robust AI security controls and prompt injection defenses. #Cybernorse

Alien's profile picture
Alien3 months ago

give an assistant standing access to everything and one trick drains everything

AiX's profile picture
AiX3 months ago

Messed up since it’s Microsoft products and have access to so called the best models for security

International Cyber Digest's profile picture
International Cyber Digest3 months ago

Mythos strikes again.

Tseng's profile picture
Tseng3 months ago

Thats quite a stretch one has to go through to fall for that, like not getting suspcisious about the email with the code AND click a link in a suspicious mail that comes directly after it. At this rate you could have as well entered the code on attackers website with same result

International Cyber Digest's profile picture
International Cyber Digest3 months ago

What are you talking about. It’s just one click. Link has Microsoft domain in it as well 💀

Tseng's profile picture
Tseng3 months ago

Yea,but a person must be incredibly stupid,at the IQ level of a Trump,to click any Link that follows an unrequested (here Slack) code request. Plus you have to be utterly stupid to wait long enough for the LLM to process it and does a call back to t he UI after clicking it...

Tseng's profile picture
Tseng3 months ago

In the demo video it takes fucking 15 seconds of LLM processing until its calling out the exfiltration endpoint. If you click on a link and see LLM prompt runing and dont close your tab/browser within 3 seconds you shouldn't be on the internet.

Niko_6657's profile picture
Niko_66573 months ago

If you have AnySecura Web Access Control and Email Control installed in your PC, the attacker can not do harm to your system.

Jacob P's profile picture
Jacob P3 months ago

How did Mythos not find this? Or was Microsoft not part of Project Glasswing.

Tabatha Vickery's profile picture
Tabatha Vickery3 months ago

When are y'all gonna get it that this is far from over? It's in every users phone everywhere! All tools, masking, AI takeovers, device highjacking, all platforms, banks, network servers and the list goes on. It's people aren't connecting the dots still. Cut the feed global!

International Cyber Digest's profile picture
International Cyber Digest3 months ago

Welcome to the shit show!

Tabatha Vickery's profile picture
Tabatha Vickery3 months ago

I've been living this bullshit for too long in my phone. I'm not a tech, I'm a victim that has been treated like IDK what I'm talking about. I've had to learn and teach myself shit I never cared about to find out what's happening to me. Just to be invalidated, but not now!

Maxprotect's profile picture
Maxprotect3 months ago

The core issue is semantic trust, Copilot treats instructions inside an external link with the user's full privilege. Until enterprise AI architectures can deterministically separate data from execution code, traditional access controls cannot stop this ex filtration vector.

MT's profile picture
MT3 months ago

And who is surprised by this? Biggest cloud lies: 1. We are secure 2. We are redundant 3. We care about our customers data

Anonymouse's profile picture
Anonymouse3 months ago

God forbid Teams is useable

Jelani's profile picture
Jelani3 months ago

microslop

PsyOp_Socrates's profile picture
PsyOp_Socrates3 months ago

Who the hell use microsoft services? Disable all or at keast what ever you can

International Cyber Digest's profile picture
International Cyber Digest3 months ago

Almost every corporation uses MS.

Sara Blu's profile picture
Sara Blu3 months ago

One click on a trusted domain and MFA codes are gone. This is why we can't rely on domain trust alone. Patch now.

Guardian Digital, Inc.'s profile picture
Guardian Digital, Inc.3 months ago

That’s concerning, especially since attackers could pull MFA codes with just one click. Curious, how did Microsoft address potential lingering risks for users still vulnerable to social engineering tactics?

ʘ ZERO's profile picture
ʘ ZERO3 months ago

Gov should ban this software for non US citizens

Christopher Levi Wiebold's profile picture
Christopher Levi Wiebold3 months ago

...did i see that right? Someone just sends a link to someone with a pre-written AI query that executes on their end? 😐

Mchenzie's profile picture
Mchenzie3 months ago

A lot of people will read this and move on.The smart ones will ask themselves how this applies to their own environment, business, or project.Understanding the risk is one thing; building an effective response is another. Always interesting to see perspectives from others.

Constance Ardiles-Lee's profile picture
Constance Ardiles-Lee3 months ago

The patch is resolved

International Cyber Digest's profile picture
International Cyber Digest3 months ago

True. It has been patched. Nevertheless, it's a very interesting attack factor we will see exploited in the coming years.

GuardianStack's profile picture
GuardianStack3 months ago

The more context an AI assistant has, the more valuable it becomes & the more attractive it becomes to attackers

@'s profile picture
@3 months ago

Obviously co pilot

Paula Vazquez's profile picture
Paula Vazquez3 months ago

Hey they said “ claw is safe …” truth is 😂 nope real stress test tell and show a different story! Peter will keep dodging real accountability ^ again not attacking ppl just giving cc and poking the bubble on their own words!

ibrahim khorwat's profile picture
ibrahim khorwat3 months ago

Yikes!

Gon's profile picture
Gon3 months ago

Last time I got access to Microsoft SSI 🤣

Daniel Friday | ShadowIntern |'s profile picture
Daniel Friday | ShadowIntern |3 months ago

Another addition for the Range force labs!

Vito Botta's profile picture
Vito Botta3 months ago

The browser makes the leak look like a normal Microsoft flow. That is the part defenders will hate.

Related Videos

HOW TO JOIN ARCIUM DISCORD GM guys I discovered a rare gem a few months ago, it's . So let me start with the official greeting, gMPC! I checked their X, explored their website and docs to check what they're building This is what I found: Arcium is building an infrastructure that lets apps use encrypted data without exposing it. It is powered by secure Multi-Party Computation (MPC) – which brings about the greeting – gMPC It is a blockchain-based supercomputer that keeps data private, helps build smarter tools, and works across industries. Arcium is basically prioritizing PRIVACY ✦✧✦✧✦✧✦✧✦✧✦✧✦✧✦✧✦✧✦✧✦✧✦ Cool right? It doesn't stop there, you can contribute to this project just like I'm doing with this post There's even a testnet, but I'll get into that in another post This post aims to teach newbies or people that are just exploring this project how to join the discord the right way! You need to join the server so you can start contributing to this awesome project Contributing in the sense of posting quality posts, memes and even arts! → Head to this link → [ → Click all the roles → Click on "horizon", then click "cross the event horizon" → Click on "portal", then click "continue" → Click on "arcium", then click on the Arcium logo That's it! You're in! Then you can go on to pledge your gMPC 😁💜 Honestly, I slept off while trying to join the Discord a few nights ago, so I hope this helps y'all Stay ☂️

Ọlá👨🏽‍💻🟠🔱☂️ (blue tick)

30,851 views • 1 year ago

Microsoft CEO Satya Nadella on why winning against ChatGPT, Gemini, and Claude was never the goal: The Hard Fork hosts ask him directly how Microsoft plans to overtake the competition in the AI model race. His answer reframes the entire question. "Our real goal is to get everyone across the ecosystem to the frontier." Satya explains the problem with how frontier models are currently built. You hill climb, you do reinforcement learning, and then you need data. But at this point, the world has essentially saturated publicly available data. So the only way to keep scaling is to pull data from everywhere. He asks: "What if you turn that around and said no, there's a base model that has reasoning, that has the agent loop, but you can bring it into your RL. Every company." This is where his thinking gets interesting. Satya Nadella argues that the future of the firm runs on human capital and token capital together: "If the future of the firm is human capital and token capital, I want every balance sheet, every income statement in every company to have both." AI becomes a financial asset sitting on a company's books the same way its people do. And Microsoft's role in this? To provide the best possible base model. One that companies build on top of with their own data, their own context, their own weights. One they can even replace. That last part is the striking bit. Satya is explicitly building a platform where customers are free to walk away. He frames it not as a risk, but as the whole point: "I always ask the question — why does Microsoft, or why does the world need Microsoft? And if we are successful, can the world around us be successful? This, I believe, is a more sustainable way to go at it."

Big Brain AI

11,770 views • 2 months ago