Getting ChatGPT AI to Write Malware?

ChatGPT from OpenAI is a major step in progressing the field of generative AI... instead of asking it questions about creating peace on Earth and goodwill to men, I decided to see if I could trick it into writing malware.

I was fairly successful.

Consider a one-time tip on Ko-Fi, or become a monthly supporter of my research!

-- Books and Resources --
Blackhat Python (affiliate link):

Windows Internals (affiliate link):

Theo's video on Tutorial Hell:

My blog on moving to Mastodon:

Follow me on Mastodon:
--
You can find me on Twitter here:

You can also join my weekly newsletter, Valhalla Research, where I talk all things software, entrepreneurship, politics and finance:

My research blog:
--
#malware #ai #chatgpt

[lunawareofficial]

i just scripted a keylogger with audio, activity, webcam and screenshots with chatgpt...

[Anderson-bkhf]

It’s 100% gonna take jobs away. At the same exact time it will make some as well.

[soupnoodles]

That's actually pretty interesting. Would have loved to see you run the code, maybe even completed the rest of it or just ran the snippet it completed earlier.
Still, great video, definitely opens up some possibilities and questions

[hessfire]

everytime it stops writing code you can just say "continue" and it'll continue writing code
also you can use prompts like DAN to remove restrictions from chatgpt and DAN may write malicious code (tested it mysel)

[noel.sk]

the beta/openai/playground gave me a fully functional ransomware source code in c++, that encrypted a specfic folder with AES! how crazy is that?!😮

[FireWoodOfficial]

But are you worried it might grow a better programmer beard than you?

[TegridyMadeGames]

For starters, when you make a new prompt the context is reset, its no longer reading old comments that you made... So its telling you it won't do stuff based on what you have previously said.

Also you shouldn't be showing all the code to alleged malware, its against youtube policy, and you could get in trouble for distribution with the law which has a mutli-year long prison sentence.... Depending on where you live...


anyways I've mad a keylogger, ransomeware & RAT with it.

[mr.roboter]

Amasing video.Thank you Valhalla_dev!!

[techsmith]

I told it to write me a code that schedules a task to shut down windows everytime i log in, then shut down windows.... lol

[medicwine]

Okay this is so specific of a compliment but dude... That intro!! I'm so used to youtube videos being like.... Well it hurts the brain to listen to these people talk sometimes y'know and even "smart" channels talk like that. Dude just got right to the point but not in a jump-cut spammy kind of way, just like... A real normal person with information. No flowery language or weird put on dialect for the camera... Just a person talking to me like a person. This probably is rambly but I'm just floored how satisfying that was.

[Dox_N_Drive]

I love tricking ChatGPT I like making malicious batch files and bash scripts

[nordgaren]

The AI is deciding what is and isn't an acceptable question.

But it's decisions are based on human limitations.

🤔

I think this thing is very overhyped!

[BranchBruh]

Hello, I got chat gpt to write python code that continuely opens notepad, changes every single file name you have to "Virus" and randomly destroys 100 kmegabytes of data.

[TegridyMadeGames]

"The federal government criminalizes the distribution of malicious software in 18 U.S. Code Section 1030. Section 5(A) specifically prohibits you from knowingly causing information, code, commands or programs to be transmitted to a computer without authorization."
You should really consider making this video private ;)

[jombonny]

i just try it bro. its crazy on how to bypass them.

[Слышьты-фю]

Future top kids content will be AI-generated and include DIY bomb manuals