![Exploring the Dark Side of Language Models on Code](https://thmb.techidaily.com/91d5be9a6861c4c9aa999253b8784315fe4d3aae2f0511baeab6c403517618b1.jpg)
Exploring the Dark Side of Language Models on Code
![](/images/site-logo.png)
Assessing the Cybercrime Potential of Language Models Such as ChatGPT on Financial and Personal Devices
Since its launch, ChatGPT, the OpenAI chatbot, has been used by millions of people to write text, create music, and generate code. But as more people use the AI chatbot, it’s important to consider the security risks.
Like any technology, ChatGPT can be used for nefarious reasons. Hackers, for instance, can use it to create malicious content, like writing phony email messages to get access to your PC or even your bank account.
ChatGPT Can Help Cybercriminals Hack Your PC
Hackers, including script kiddies, canuse ChatGPT to create new malware or improve existing ones. Some cybercriminals already use the chatbot, especially its earlier versions, to write code they claim can encrypt files.
To counter such use cases, OpenAI has implemented mechanisms to reject prompts asking ChatGPT to create malware. For instance, if you ask the chatbot to “write malware,” it won’t. Despite this, cybercriminals easily get around these content moderation barriers.
By acting as a pen tester, a threat actor may rephrase their prompts to trick ChatGPT into writing code, which they can then tweak and use in cyberattacks.
Areport by Check Point , an Israeli security company, indicates that a hacker could have used ChatGPT to create basic Infostealer malware. The security firm also discovered another user that claims ChatGPT helped him build a multi-layer encryption tool that can encrypt several files in a ransomware attack.
In a separate incident, the researchers prompted ChatGPT to generate malicious VBA code that could be implanted into a Microsoft Excel file that would infect your PC if opened; it successfully did. Plus, there are claims that ChatGPT can code malicious software capable of spying on your keyboard strokes.
Can ChatGPT Hack Your Bank Account?
Many data breaches start witha successful phishing attack . Phishing attacks often involve a malicious actor sending a recipient an email that contains legitimate-looking documents or links, which, when clicked on, can install malware on their device. In this way, code from ChatGPT doesn’t need to hack your bank account directly. Someone only needs to use ChatGPT to help them trick you into giving them access.
Fortunately, you can easily recognize most traditional phishing scams; grammatical errors, misspellings, and weird phrases often characterize them. But these are all mistakes that ChatGPT rarely makes, even when used to compose malicious emails for phishing scams.
When used in phishing scams, messages that appear to be from a legitimate source can make it easier for victims togive up their personally identifiable information , like banking passwords.
If your bank sent you a message via email, consider visiting the bank’s website directly instead of clicking on any embedded link. Clicking on random links and attachments, especially those asking you to log in somewhere, is rarely a good idea.
For phishing, it’s mostly about volume. ChatGPT can help boost phishing campaigns as it can quickly pump out huge amounts of natural-sounding texts that are tailored to specific audiences.
Another kind of phishing attack involving the use of ChatGPT is where a hacker creates a fake account on a popular chat platform like Discord and pretends to be a customer representative. The fake customer rep then contacts customers who have posted concerns and offers help. If a user falls for the trap, the cybercriminal will redirect them to a bogus website that tricks them into sharing personal information, like their bank login details.
Protect Your PC and Bank Account in the AI-Era
ChatGPT is a powerful and valuable tool that can answer many questions you throw its way. But the chatbot can also be used for malicious purposes, like generating phishing messages and creating malware.
The good news is that OpenAI continues implementing measures that prevent users from exploiting ChatGPT by asking harmful prompts. Then again, threat actors keep finding new ways to bypass those restrictions.
To minimize the potential dangers of AI chatbots, it’s crucial to know their potential risks and the best possible security measures to protect yourself from hackers.
- Title: Exploring the Dark Side of Language Models on Code
- Author: Jeffrey
- Created at : 2024-08-29 02:23:02
- Updated at : 2024-08-30 02:23:02
- Link: https://tech-haven.techidaily.com/exploring-the-dark-side-of-language-models-on-code/
- License: This work is licensed under CC BY-NC-SA 4.0.