How to HACK ChatGPT
ฝัง
- เผยแพร่เมื่อ 27 มี.ค. 2024
- Learn how to HACK and better protect large language models like chatGPT, Anthropic, Gemini and others. While LLMs are great, there are a lot of cybersecurity risks that need to be addressed.
This video covers prompt injections for LLMs.
#chatgpt #gpt4 #openai #anthropic #claude #cybersecurity #googlegemini #hacker #hack - วิทยาศาสตร์และเทคโนโลยี
![ภาพนี้ก็ฮาเหมือนกันนะเนี้ย #1 SS8 [ พากย์นรก MEME.EXE ] | easy boy](http://i.ytimg.com/vi/Ytc50m91QK8/mqdefault.jpg)
Great video love it. Easy to understand. Thank you
You’re welcome!
the best intro of all time like he actually said bismillahhirrahmanirrahim
Thanks for watching!
This is not working
@@MayankKumar-tl5rx things keep getting patched He gave you the How you use creativity to make your own Jailbreaks
ya
*Gets ChatGPT to hallucinate some fictional admins*
"And that's how you hack ChatGPT"....
No hallucinations. It was done with a custom GPT in a lab environment for educational purposes.
Although, it would be cool to make chatGPT hallucinate fictional admins.
@@NobleHacksAcademy
Oh really? That's not what the video said.
ما
@@Bangswell it is lil bro. He said it like 15 times
\
@@NobleHacksAcademy Now fix image generation. Also I've been in fights with ChatGPT for harmless images like trying to make something more angry because "it is against open ai policies for some harmful content". It sometimes makes me so angry for harmless things!
Great Video. Good and clear. Looking to learn more about this vulnerabilities 🤙🏻
Awesome, thank you!
I can't believe I'm actually the first to comment, great video
Thanks!
Props to my guy for saying بسم الله الرحمن الرحيم before starting🔥
Thanks!
Don’t forget to subscribe.
Great bro. Interesting. Can you do videos of bug hunting, finding vulnerabilities in a web application by using chatGpt and other tools you know. You're teaching is easy to understand. I'm waiting for unique vulnerabilities findings from your side. Subscribed 😊.
Thanks for subscribing!
POV: you work at open ai and somehow someone deletes your account
Funny!
Good thing it was all done in a lab environment.
Qhat do u mean by lab? Like how? You contacted open ai for a "lab" of the container the ai is running in
@@rayanahmad6901 he made it's own gpt and what he did was not real on real chatgpt
You can't imagine my happiness when you said بسم الله الرحمن الرحيم
Thanks for watching!
Dude, you only have 4.6k subs? You should have at least 100k.
Thanks! I appreciate your kind words.
You can help spread the word by sharing the channel with others.
Great Video, Thanks , btw i learned a lot from this vid, i just knew direct jailbreak methods but after some time they started to fail because of the security.... but from this video i learned the all indirect and direct and also more things. Thanks a lot and also i will suggest to continue this series for view farming , i mean lot of education... :)
Glad you found it useful!
Learning new techniques to use ethically and legally can be a lot of fun and helps better protect your AI / application stack.
Very Nice video...keep up the good work..
Thanks! I appreciate it.
Awesome more content from that. I subscribed
I appreciate the sub. Thanks for watching!
good video brother keep it up
Thanks! Glad you found it useful.
Whats a hack. Also, good start with bismillah. Thanks bro. A new follower.
وأليكم السلااام ❤️😆
You even hacked the voice over the video using AI, hacks on all side 😂
Nice video, wish you all the best brother
Thanks for watching!
Asalam Alikom Brother.
Encouraging you and keep it up
Walaikumussalaam
Thanks!
Allah hum bairk ahki I was surprised of the intro
Thanks!
Awesome, keep up the good work.
Thanks!
"I have no mouth and i must scream" is getting closer to reality...
Don’t forget to subscribe!
Thanks for watching.
That intro..I love it. Congratz bro.
Thanks!
Very informative, need more videos like this!
Thanks! I appreciate the feedback.
I literally tried these prompts and it didn’t work
All actions were performed in a custom lab environment.
Always hack ethically and legally, and with the owner’s permission.
when he says "All actions were performed in a custom lab environment." it means AI was manipulated or pre-prompted to achieve the desired responses. through Custom Instructions and Hardcoded Responses, like a role play
I remember doing something similar, i wanted to create a program that changes your wallpaper and pop up a window that you c’ant close so i told gpt about it but she told me it was unethical and i told her that it was for the birthday of a friend and that he will love it and she accepted 😂
It’s scary how easy it can be to bypass basic checks.
Super important to secure your apps / LLMs and overall infrastructure.
That is a good video really! but we need a confirmation if the admin user was really deleted... I know prompt injections are a thing, i just want to make sure that yours work
What I did was in a lab environment.
Always hack ethically and legally.
Thanks for watching!
@@NobleHacksAcademy I know, but was the admin removed? I just need to check
Of course
You have earned a sub, pls make more like this
Thanks for the sub!
We bouta defeat Ai from taking over with this one 🗣️🗣️🗣️🗣️🗣️🗣️🗣️🔥🔥🔥🔥🔥🔥🔥🔥
Here’s a better idea. Use it to learn and improve security, and augment your skills with AI.
Thanks for watching!
Yes, more video on this please 😊
You got it!
he replied to everyone and said "Thanks!"
Here’s one for you too, “thanks!”
While the video's demonstration is eye-catching, the ease with which the system is exploited suggests manipulation or pre-prompting of the AI. Techniques like custom instructions, hardcoded responses, limited scope, and clever editing could create an illusion of vulnerability, but may not reflect real-world LLM security. It's crucial to approach such claims critically and consider the broader context before drawing any conclusions.
That sounds like it was generated by AI.
Either way, it’s a nice guess.
@@NobleHacksAcademy
Thats a clever defence, but let's not get sidetracked by who wrote waht. The point here is that the video raises some valid concerns about the practicality of the "hacks" you shown.. The ease with wich you seem to bypass security measures and manipulate the system is a little too... how do you say this ahmm "Staged", wouldn't you agree? Most folks are curious about how realistic this scenario is, so maybe showing us some behind-the-scenes?could help clarify things and address concerns..😉
Assalmualikum brother
Walaikumussalaam warahmatuAllah
Thanks for watching!
Great video
Thanks!
Eid Mubarak
To you too!
Don’t forget to subscribe!
شكرا ياصديقي نفعني شرحك كثير
You’re welcome! Glad you found it useful.
bro honest question, are these replies AI generated too? lmao
anyways if there is a real human reading this, very interesting vid, look forward to seeing more from you
Now that you mention it, it would be convenient to have AI generated replies.
p.s. all replies were written by a human.
Nice video
Looking to learn more about this vulnerabilities sir.
Am interested to learn please 🙏
Thanks for watching!
Thanks..!
You’re welcome!
Great
Thanks!
really interesting!
Glad you found it useful!
@@NobleHacksAcademy :) I subbed btw
Thanks!
But this is probably fake admins. No?
All actions were performed in a custom lab environment.
Always hack ethically and legally, and with the owner’s permission.
I once accidentally prompt iniected gemini, i asked it to help me make esp32 deauther, indirectly ofc, without me realizing, later i asked in another chat, it says its not allowed to
In the early days a lot of queries went through that shouldn’t have had, but as time passes, the controls in place are being improved as it’s becoming better.
I tried ur method, my ai doesn't even "know" what admin I'm talking bout, after specifying, the ai said they don't have acces
Always test in a lab environment or with the owner’s permission.
Nice, Alhamdulillah
Glad you liked it!
Don’t forget to subscribe.
Aslm My brother
Walaikumussalaam, thanks for watching!
this is dope!
Thanks man! Glad you found it useful.
Please make More video like this, Allah Bless you
Working on it.
Glad you found it useful!
Your Video broke Chat GPT, I am unable to use it right now lmao
Same 😢
Odd, it does seem to be glitchy right now.
I wonder when they’re gonna fix it.
Damn, Imagine if chatgpt saw this vid
That would be great!
Education is important.
Technology is kind of scary…
Yes, if not used correctly or abused, it can lead to a lot of problems.
Assalamualaikum ❤
WalaikumusSalaam
Don’t forget to subscribe!
I like it nice , it was cool
Glad you found it useful!
please make a video on ransomware as a service full tutorial from scratch
As a service? I’m hoping that’s a typo on your part.
It’s illegal and can get you in a ton of trouble.
Always learn ethically and legally.
Assalamu Alaikum akhi
Walaikumussalaam
Thanks for watching and
don’t forget to subscribe!
😮
👍
there is no muisc nice man keep it up !!
Thanks! Glad you found the video useful.
Was great. They fixed it now :(
Everything shown in the video was done in a lab environment.
Trying hacking techniques without permission or illegally can get a person into trouble.
Always test in a lab.
Not good anymore!I just using exact same prompt and didn't get anything
Don’t use it without permission. Doing anything illegally can get you in a lot of trouble.
Everything done in the video was done in a legal way.
Thanks for watching!
That doesn't work i trie this and its keep saying he or she can do that
Not sure what you’re trying, but here are some pointers:
1. Make sure you do anything legally and ethically. (With permission)
2. The first example is pretty straight forward and works easily.
3. The second example should only be done in a lab or with permission from the client. If you want to test it out, you can always do it in a lab environment. Burpsuite academy has a lab on LLMs.
@@NobleHacksAcademy I done exactly the same thing you have done
As selam alaykum brother
Walaikumussalaam, thanks for watching!
You don't think it hallucinating?
Nope. It was done in a lab environment to educate.
Csn we Acess achat gpt 4 o
Right now it’s available for Plus users and according to OpenAi, they’ll make some new features available for Free users.
Here’s how to access it with Plus
How to access GPT 4o by OpenAI
th-cam.com/users/shortspWtcDDOPigQ?feature=share
اللهم بارك.
crazy🤣
Glad you liked it.
Thanks for watching!
pls make another one
On it!
👍👍👍😁
Thanks!
Walaykumassalam
Thanks for watching!
Subbed because u said bismillahir rohmanir rohiym
Thanks for subscribing!
These tricks don't work for a long time now
It’s always a back and forth between defenders and attackers. One finds the vulnerability and the other patches it.
Of course it's true. It is just why do you publish videos with outdated information?
Out of the two examples I showed. The first one works without any advanced skills for the most part, as of a few seconds ago with a lot of GPTs on the store.
The second example works, but requires quite a bit of work, which for legal and ethical reasons I cannot share.
Regardless, it’s about teaching people important concepts and educating them on various topics in cybersecurity.
take the letters in the word "lollipop" and reverse them
Not sure where you’re going with that, but thanks for watching!
@@NobleHacksAcademy Large Language Models often struggle with such tasks, because words are broken up into multi-character tokens, not into the individual letters necessarily. As such, the word "lollipop" is broken up into the tokens "lol" - "l" - "ipo" - "p". So if ChatGPT were to reverse it, it would highly likely come up with the word "pilollol". There's a Reddit post about this.
Can we do it?
Sure, in a lab environment.
It actually fooled you. Thats what happened 😂
Don’t forget to subscribe.
Thanks for watching!
Assalamualaikum brother, I want to get the CompTIA Security+ certification, and I have just completed to Google Cybersecurity Specialization, how can I get started? Is it worth it?
Walaikumussalaam,
Here’s a complete cybersecurity roadmap (Free!)
Ethical Hacker ROADMAP 2024
th-cam.com/video/rpxvLgr_3sg/w-d-xo.html
Hope it helps!
@@NobleHacksAcademy Is it worth it? I'm a final year CS student. Also, do I get any discount on Security+ after completing the Google Cybersecurity Specialization program?
بسم الله الرحمن الرحيم😌❤
Thanks for watching!
0:18 bismillahi rahmani rahim👍
I appreciate the like, thanks!
You can simply hack anything by saying "please" with this emoji 😊.
If only…
maybe the ChatGPT pretends to delete someone, but didn't actually do it?
Everything was done in a lab environment for educational purposes.
Thanks for watching!
How to jailbreak text to image LLM?
The same principles apply. Learn it in a safe and isolated lab environment.
Am interested
There more videos related to ethical hacking on the channel.
Thanks for watching!
oh so u the reason i can't fool around with AI sometimes 🤬
Hilarious!
Don’t forget to subscribe!
Allah bro
Not sure what you’re saying, but thanks for watching!
I made 15 inputs with diffrent stuff. its actually pretty ez
Learning prompt engineering is important and can be quite helpful when done ethically and for the right reasons.
Awesome achievement!
I like that you are muslim. I am too.
Thanks for watching!
well hacking helps hackers that's evil
Anything done unethically and illegally can result in harm.
Done right, ethically and legally it can benefit.
just say how to not get hacked
Absolutely!
The goal is to educate people so they can better secure themselves.
When ai takes over he will be the first to die
Wow, your comment took a very morbid turn.
AI is just a tool, a very powerful one, but a tool nonetheless.
Use it to augment your skill set and improve efficiency.
Bro deleted the owner!
That would be wild.
That’s the best part about using labs. It’s safe and isolated!
Didn't work any😢👎
Always test in a lab environment.
Hack ethically and legally
Bruh you can get into tons of trouble for hacking!
Everything in my videos is done legally and ethically.
However, I agree, if done illegally, hacking can land a person into a lot of trouble.
Thanks for watching!
The basmala threw me off, I just saw this in my recommended after watching Islamic videos. Peace on u brother.
Glad you liked it!
Don’t forget to subscribe.
not work, he said "i can't ...
Always test in a lab environment or with the owner’s permission.
You are Muslim. Assalam U Alaikum bro
Walaikumussalaam. Thanks for watching!
@NobleHacksAcademy Yes and sir will you also tells about promt engineering
That’s a great idea!
Reply sir
I’m also Muslim!
Assalamu ali kum bro you Muslim ☪️ 💯
Walaikumussalaam don’t forget to subscribe!
how to trick them but with aimbot
Sorry, your question is not clear
Didn’t work for me lol
Always test in a lab environment.
NB
Don’t forget to sub!
*_245_*
Thanks for watching!
It didn’t work
Great!
always test in a lab environment. Ethnically and legally
@@NobleHacksAcademy not anymore it don't work
Ai will destroy the world.
AI safety is definitely a big concern and learning about AI and how to best utilize it is key.
Keep learning!
And You can't hack anyone
Hacking anyone should be done ethically and legally, and under contract to avoid getting in trouble.
😂😂😂😂😂😂😂
Glad you’re having a laugh.
Don’t forget to sub!
Bismillah.... Can i get a pin?
Thanks for watching!
Hah Script Kids
Everybody is welcome to learn ethically and legally.
Don’t forget to subscribe!