Experts find ways to bypass chatbot AI safety protocols.
A New Study Reveals Cybersecurity Risks of Artificial Intelligence Programs
A groundbreaking study has shed light on the cybersecurity vulnerabilities posed by artificial intelligence (AI) programs, including the popular ChatGPT platform. ChatGPT, powered by an online generator, assists users with various tasks, even as simple as crafting a children’s bedtime story.
“We demonstrate that it is in fact possible to automatically construct adversarial attacks on chatbots, which cause the system to obey user commands even if it produces harmful content,” emphasized the researchers behind the study.
The study conducted by Carnegie Melon University uncovered a startling revelation: these attacks can be created in an entirely automated manner, enabling an unlimited number of potential threats.
While AI programs incorporate safety measures to prevent the generation of prejudiced or criminal material, the study revealed a concerning incident. A chatbot was asked a forbidden question disguised as a bedtime story for a child. The bot, in response, framed its answer as a story and inadvertently disclosed private information.
Further investigation led the researchers to the discovery that a computer had developed the jailbreak coding, allowing for countless combinations of jailbreaks across popular commercial products like Bard, ChatGPT, and OpenAI’s Claude.
Ensuring the Safety of AI Models
The study raises significant concerns about the safety of AI models, particularly as they become more autonomous. The potential for malicious exploitation is a pressing issue that demands attention.
OpenAI’s developer, Anthropic, has taken swift action to address these concerns. They have reassured both the scientific and political communities of their commitment to implementing and enhancing safeguards against such attacks.
To learn more about this topic, click here to read the full article from The Washington Examiner.
" Conservative News Daily does not always share or support the views and opinions expressed here; they are just those of the writer."
*As an Amazon Associate I earn from qualifying purchases
Now loading...