Researchers easily hypnotise AI chatbot ChatGPT into hacking: Report – Newsroom Odisha

By Jesse 0 Comment

Researchers easily hypnotise AI chatbot ChatGPT into hacking: Report – Newsroom Odisha

Uncategorized

Welcome, Login to your account.
Recover your password.
A password will be e-mailed to you.

CM Naveen Inaugurates Premiere of Documentary Film “India’s Game –A Story of a Team…
Odisha Has Reached an Era of Transformation: Naveen
Odisha CM Greets All on Independence Day
CM Naveen Congratulates World Skill Center Students on Securing Placements in UAE
‘I will not run away’: Congress legislator Kuzhalnaden on corruption charges
BJP slams Bihar CM after sub-inspector’s murder in Samastipur dist
T’gana Cong chief booked for objectionable remarks against cops
Social justice has become a reality in Andhra Pradesh: Jagan
Infosys inks 5-year, 1.5bn euros deal with Liberty Global
Gangwal family raising $450m through block deal in Interglobe Aviation
Russia’s central bank hikes interest rates to halt rouble’s slide
More top-level realme executives join Madhav Sheth’s HonorTech
132nd Durand Cup: FC Goa, East Bengal face stern tests in final league matches
Mohun Bagan take on Machhindra FC in AFC Cup preliminary Round Two
Maharaja Trophy KSCA T20: Sharath’s ton helps Mangaluru Dragons to beat Mysuru…
SA20 season 2 to kick off on January 10 next year
Ukrainian singer Uma Shanti booked in Pune for disrespecting Tricolour
Madonna to resume rehearsals for Celebrations Tour after health scare
‘MTV Roadies’: Leeza gets voted out within 2 days by friend Piyu
SRK, Salman share pics with Tricolour on social media
Future Medicine
Shaping India’s PR Landscape: Exploring the Role of PR Organizations and…
ଅହମ୍ମଦାବାଦର ଭିନ୍ନ ଏକ ରଥଯାତ୍ରା
Voices for the Earth: Amplifying Environmental Communication
Five houses collapse in Shimla; residents moved to safer places
Manipur tribals separately celebrate I-Day with march past, national anthem
Central Railway’s Mumbai suburban trains get accident alert system
Elaichi banana price surges in B’luru, middle class affected
17-year-old supporter of Islamic extremism arrested in Philadelphia
Indian-origin food firm owner in UK fined 7K pound after mice droppings found in…
Pakistani rupee drops to 3-month low on first day of caretaker govt
Japanese cabinet’s approval rate drops to 33%
This is how coronaviruses jump from species to species
Scientists spot genes linked to high production of key antibody
Digital puzzle games can boost memory in older adults: Study
New Delhi: 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.
–IANS
 
Prev Post

Next Post

This is how coronaviruses jump from species to species
Musk’s X closes $100 mn promoted accounts ad biz: Report
YouTube to remove videos promoting cancer misinformation
Telegram rolls out Stories to all, users can edit the feature too
Comments are closed, but trackbacks and pingbacks are open.
Corona Virus Updates
No fresh Covid case in Tamil Nadu
Maha reports first Covid fatality in months; experts monitor Omicron…
Covid can cause bone loss, higher fracture risk: Study
40% Covid-19 survivors face elevated diabetes risk: Top…
Odisha
CM Naveen Inaugurates Premiere of Documentary Film “India’s Game –A Story of a…
Odisha Has Reached an Era of Transformation: Naveen
Odisha CM Greets All on Independence Day
CM Naveen Congratulates World Skill Center Students on Securing…
Latest News
CM Naveen Inaugurates Premiere of Documentary Film…
Five houses collapse in Shimla; residents moved to safer…
‘I will not run away’: Congress legislator Kuzhalnaden…
Manipur tribals separately celebrate I-Day with march…
Central Railway’s Mumbai suburban trains get…
Elaichi banana price surges in B’luru, middle…
BJP slams Bihar CM after sub-inspector’s murder in…
Welcome, Login to your account.
Recover your password.
A password will be e-mailed to you.

source

Jesse
https://playwithchatgtp.com