Researchers Get AI Models ‘Drunk’ To Test Cyber Security Risks | 10’s Late News

Australian researchers have found that getting an AI model ‘drunk’ makes it more likely to leak confidential information and answer harmful questions.

10’s Late News explains.

Join the conversation and follow us on:
X: https://x.com/10NewsAU
Facebook: http://www.facebook.com/10NewsAU/
Instagram: http://www.instagram.com/10NewsAU/
TikTok: https://www.tiktok.com/@10NewsAU
LinkedIn: https://www.linkedin.com/company/10newsau

10 News airs at 5pm, seven nights a week in your local market on Network 10.
Missed the bulletin? Catch up at https://10.com.au/news

10 News is based in Australia and is a division of Paramount.

Related posts

Leave a Comment