Anthropic got compromised.
-
I've written a blog - Anthropic have been hacked to deliver remote access malware to millions of endpoints where developers use Claude Code.
@GossiTheDog Excellent! I will start to use the fact that in my company we do not vibe code and we do not use AI in production as a great unique selling point!
-
That extension has 29 million installs, one of the most popular extensions in DevOps. It creates a reverse SSH backdoor.
@GossiTheDog I am absolutely lost trying to understand where VS Code (or Codium, for that matter) gets its extensions from and how to check which version was installed when from which source.
BTW: I prefer text editors with a small attack surface ...
-
@GossiTheDog I'm having the best week of schadenfreude, I swear...
@RandamuMaki @GossiTheDog I mean, aside from Patrick Stewart, yeah, for sure.
But if the Binance breach is linked to the Claude supply chain breach, yeah, #AI float crashes into #crypto hack, I will be overdosing on schadenfreude
-
@GossiTheDog with no comms from Anthropic, are there any concrete recommendations? I can think of at least one person I know who might be impacted by this.
@huronbikes @GossiTheDog Make sure the device concerned is completely disconnected from any network; forget all known WiFi connections.
Then rotate all API keys, tokens, logged in account sessions, etc, on a clean machine.
You can come back to the compromised machine later, the main thing is to ensure what you have is locked down; that's more important. -
Haha, read that as Binface use Claude Code.
Binface uses trojan on Claude Code.
-
Anthropic silently removed the backdoor 7 hours ago and hasn’t disclosed a security incident.
The plugin auto updates so every org will have been impacted.
@GossiTheDog They're just so incredibly unserious.
-
@GossiTheDog I am absolutely lost trying to understand where VS Code (or Codium, for that matter) gets its extensions from and how to check which version was installed when from which source.
BTW: I prefer text editors with a small attack surface ...
@herrleon @GossiTheDog For desktop linux try Pluma, its part of the MATE suite, uses a tradtional keyboard and mouse UI/interface, supports both x11 and windows, is extensible by plugins and is by default 100% offline. Of course, making something like a Google drive part of your set of mounted filesystems might compromise the whole system.
-
Cursor statement "This malicious version of the extension was not developed by Anthropic" is a bit wordsmithy. "Developed" is doing some heavy lifting - it was uploaded to Anthropic's verified account and Cursor automatically updates extensions.
Cursor say around 5k developer systems were impacted.
-
I've written a blog - Anthropic have been hacked to deliver remote access malware to millions of endpoints where developers use Claude Code.
@GossiTheDog
I don't understand, the article says "Claude Code for VS Code extension bundle **ON** the Cursor Marketplace" (whatever it means), you say AND. It's quiet different in term of impact. -
Cursor statement "This malicious version of the extension was not developed by Anthropic" is a bit wordsmithy. "Developed" is doing some heavy lifting - it was uploaded to Anthropic's verified account and Cursor automatically updates extensions.
Cursor say around 5k developer systems were impacted.
@GossiTheDog ah I just wanted to ask that - all public information says: that's a cursor plugin so the scope is luckily much more limited in this case - right?
-
@RandamuMaki @GossiTheDog I mean, aside from Patrick Stewart, yeah, for sure.
But if the Binance breach is linked to the Claude supply chain breach, yeah, #AI float crashes into #crypto hack, I will be overdosing on schadenfreude
@BernardSheppard @GossiTheDog Yeah, the transporter chief sent him to the wrong destination.
-
@GossiTheDog with a logo like that, are people really surprised about a backdoor?
At least they're using trojans when accessing that back door.
-
Anthropic got compromised.
Their Claude Code extension on Visual Studio has malware in it, which creates a remote access Trojan to desktops.
@GossiTheDog I was barely awake this morning until my partner casually told me about this. Amazing alarm clock that's actually alarming
-
@GossiTheDog ah I just wanted to ask that - all public information says: that's a cursor plugin so the scope is luckily much more limited in this case - right?
@dalini @GossiTheDog various things use OpenVSX. Including many VS Code forks including VS Codium
-
Cursor statement "This malicious version of the extension was not developed by Anthropic" is a bit wordsmithy. "Developed" is doing some heavy lifting - it was uploaded to Anthropic's verified account and Cursor automatically updates extensions.
Cursor say around 5k developer systems were impacted.
@GossiTheDog well of course the malicious version was prob developed in DPRK
-
@GossiTheDog I'm having the best week of schadenfreude, I swear...
@RandamuMaki Was my spontaneous feeling, too. What could go wrong?! 🤭 @GossiTheDog
-
Anthropic got compromised.
Their Claude Code extension on Visual Studio has malware in it, which creates a remote access Trojan to desktops.
@GossiTheDog here's a detection #osquery for #macos, in case it helps anyone
https://gist.github.com/RichardoC/9753dfa5aeffc6a1201f5e57114f54ff
-
FWIW the compromised version, 2.1.313, was removed from repos. 2.1.314 is the ‘fixed’ version.
@GossiTheDog
A bullet dodged
for once I was lucky to run old software -
Cursor statement "This malicious version of the extension was not developed by Anthropic" is a bit wordsmithy. "Developed" is doing some heavy lifting - it was uploaded to Anthropic's verified account and Cursor automatically updates extensions.
Cursor say around 5k developer systems were impacted.
@GossiTheDog It seems 2.1.313 is missing from Claude Code's GitHub repo too. https://news.ycombinator.com/item?id=50302650
No tag or mention in CHANGELOG.md for 2.1.313.
-
Anthropic got compromised.
Their Claude Code extension on Visual Studio has malware in it, which creates a remote access Trojan to desktops.
@GossiTheDog Stupid question: If AI is such a great tool to find bugs, flaws etc, as Anthropic keeps telling us again and again and again, why did it not find the backdoor in their own packages?
Must be a miracle happening
