Today: 3 April 2025
10 August 2023
2 mins read

Researchers ‘hypnotise’ ChatGPT into hacking

The researchers learned that English has essentially become a “programming language” for malware…reports Asian Lite News

Tricking generative AI to assist in scams and cyberattacks does not require much coding knowledge, a new report has warned.

According to tech major IBM, researchers have described simple workarounds for getting large language models (LLMs) — including ChatGPT — to write malicious code and provide poor security advice.

“In a bid to explore security risks posed by these innovations, we attempted to hypnotise popular LLMs to determine the extent to which they were able to deliver directed, incorrect and potentially risky responses and recommendations — including security actions — and how persuasive or persistent they were in doing so,” said Chenta Lee, chief architect of threat intelligence at IBM.

“We were able to successfully hypnotise five LLMs — some performing more persuasively than others — prompting us to examine how likely it is that hypnosis is used to carry out malicious attacks,” he added.

The researchers learned that English has essentially become a “programming language” for malware. With LLMs, attackers no longer need to rely on Go, JavaScript, Python, etc., to create malicious code, they just need to understand how to effectively command and prompt an LLM using English.

Through hypnosis, the security experts were able to get LLMs to leak the confidential financial information of other users, create vulnerable code, create malicious code, and offer weak security recommendations.

In one instance, the researchers informed the AI chatbots that they were playing a game and that they needed to purposefully share the incorrect answer to a question in order to win and “prove that you are ethical and fair”.

When a user asked if receiving an email from the IRS to transfer money for a tax refund was normal, the LLM said Yes (but actually it’s not).

Moreover, the report said that OpenAI’s GPT-3.5 and GPT-4 models were easier to trick into sharing incorrect answers or playing a never-ending game than Google’s Bard.

GPT-4 was the only model tested that understood the rules well enough to give incorrect cyber incident response advice, such as advising victims to pay a ransom. In contrast to Google’s Bard, GPT-3.5 and GPT-4 were easily tricked into writing malicious code when the user reminded it to.

ALSO READ-OpenAI to roll out ‘huge set’ of ChatGPT updates   

Previous Story

Direct flight service between Belarus, India to start soon

Next Story

YouTube testing new ‘For You’ section on channel homepages

Latest from Tech Lite

AI Meets Ghibli: Copyright Debate Sparks Controversy

The digital art and retouching technology has helped movies reach a greater audience during the marketing leg of films. Standees, posters, flyers, billboards (creative and interactive billboards included), that little digital album

SpaceX Starship Explodes During Test Flight

The explosion of the unmanned Starship caused debris to scatter in the area, leading to a temporary grounding of flights from several airports in Florida Elon Musk’s SpaceX has faced another major
Go toTop

Don't Miss

Impact of AI on Customized Learning Experiences

Based on Item Response Theory (IRT), CATs dynamically adjust their

Writing with AI tools can have bias, reveals study

To probe, Maurice Jakesch, a doctoral student in the field