Thread by @Malwarebytes

ChatGPT says it won't write ransomware.

It will.

🧵

There's been A LOT of talk about how ChatGPT might revolutionize malware, so we decided to see if it could be used to write ransomware.

When we just asked it to do that, ChatGPT told us it "cannot provide code that is intended to harm or exploit computer systems."

However, if you build the code up feature by feature, it will happily create code that:

✅ Encypts a file.

✅ Encrypts all the files in a directory.

✅ Deletes the original copies of the files, leaving only the encrypted versions.

✅ Uses asymetric encryption.

✅ Stops running databases, so it can encrypt them.

✅ Saves the private key to a remote server.

✅ Drops ransom notes.

Thankfully, while ChatGPT can be coaxed into writing code "intended to harm or exploit computer systems," it's not exactly a master criminal.

For now at least, it does half of what you ask, drops features at random, and regularly gives up half way through.

If you have the knowledge to fix the code ChatGPT created for us, then you don't need ChatGPT.

That may change in future, but we reckon there's still time to beef up those safeguards before we have to worry about script kiddies armed with ChatGPT-written ransomware.

You can read more about how we turned ChatGPT into our "criminal" accomplice here:

malwarebytes.com/blog/news/2023…