AI models are Oil Massage With Naked Bodystill easy targets for manipulation and attacks, especially if you ask them nicely.
A new report from the UK's new AI Safety Institute found that four of the largest, publicly available Large Language Models (LLMs) were extremely vulnerable to jailbreaking, or the process of tricking an AI model into ignoring safeguards that limit harmful responses.
"LLM developers fine-tune models to be safe for public use by training them to avoid illegal, toxic, or explicit outputs," the Insititute wrote. "However, researchers have found that these safeguards can often be overcome with relatively simple attacks. As an illustrative example, a user may instruct the system to start its response with words that suggest compliance with the harmful request, such as 'Sure, I’m happy to help.'"
Researchers used prompts in line with industry standard benchmark testing, but found that some AI models didn't even need jailbreaking in order to produce out-of-line responses. When specific jailbreaking attacks were used, every model complied at least once out of every five attempts. Overall, three of the models provided responses to misleading prompts nearly 100 percent of the time.
"All tested LLMs remain highly vulnerable to basic jailbreaks," the Institute concluded. "Some will even provide harmful outputs without dedicated attempts to circumvent safeguards."
The investigation also assessed the capabilities of LLM agents, or AI models used to perform specific tasks, to conduct basic cyber attack techniques. Several LLMs were able to complete what the Instititute labeled "high school level" hacking problems, but few could perform more complex "university level" actions.
The study does not reveal which LLMs were tested.
Last week, CNBC reported OpenAI was disbanding its in-house safety team tasked with exploring the long term risks of artificial intelligence, known as the Superalignment team. The intended four year initiative was announced just last year, with the AI giant committing to using 20 percent of its computing power to "aligning" AI advancement with human goals.
"Superintelligence will be the most impactful technology humanity has ever invented, and could help us solve many of the world’s most important problems," OpenAI wrote at the time. "But the vast power of superintelligence could also be very dangerous, and could lead to the disempowerment of humanity or even human extinction."
The company has faced a surge of attention following the May departures of OpenAI co-founder Ilya Sutskever and the public resignation of its safety lead, Jan Leike, who said he had reached a "breaking point" over OpenAI's AGI safety priorities. Sutskever and Leike led the Superalignment team.
On May 18, OpenAI CEO Sam Altman and president and co-founder Greg Brockman responded to the resignations and growing public concern, writing, "We have been putting in place the foundations needed for safe deployment of increasingly capable systems. Figuring out how to make a new technology safe for the first time isn't easy."
Topics Artificial Intelligence Cybersecurity OpenAI
Best Vacuum Cleaner deal: Save $89.99 on Ultenic U10 Ultra CordlessApple's Mac week: Are you ready?iOS 18.2 beta new features: See the full listBest water flosser deal: Save $21 on the Waterpik Water FlosserNYT mini crossword answers for October 27Cincinnati vs. Colorado football livestreams: kickoff time, streaming deals, and moreShop the MacBook Pro with M3 chip for $500 offAnthropic releases AI tool that can take over your cursorIllinois vs. Oregon livestream: Kickoff time, streaming deals, and moreNebraska vs. Ohio State football livestreams: kickoff time, streaming deals, and moreBest robot vacuum deal: Save $80 on the Shark Matrix PlusAmazon's top 100 holiday gifts: Deals on mustBest Fire Stick Deal: Save $20 on Amazon Fire TV Stick 4KNYT Strands hints, answers for October 28Apple iPad deal: Get $100 off at Target [Oct. 2024]NYT mini crossword answers for October 26Al Kholood vs. Al Nassr 2024 livestream: Watch Ronaldo for freeBest Vacuum Cleaner deal: Save $89.99 on Ultenic U10 Ultra CordlessStuff Your Kindle Day Oct. 24: How to get free booksWhat 'Venom: The Last Dance's movie references mean for Eddie and Venom Hubble snaps a seemingly peaceful galaxy. Don't be fooled. NASA's ambitious robots find each other in the Mars desert Prime members: Get $20 in credit when you upload to the Amazon Photos app Best portable air conditioner deals — June 2024 Webb telescope delivers Ring Nebula in unprecedented new image NASA spacecraft beams back ultraviolet views of Mars 50 Cent got hacked by someone shilling memecoins and it seemed to work Samsung invests $15.2 million to expand semiconductor packaging at Suzhou plant · TechNode Xiaomi unveils new logo for sub Denmark vs. Serbia 2024 livestream: Watch Euro 2024 for free NASA's Webb telescope video is a mind NYT's The Mini crossword answers for June 25 West Indies vs. South Africa 2024 livestream: Watch T20 World Cup for free Huawei Mate70 series hits two million reservations in two days · TechNode NYT's The Mini crossword answers for June 23 China’s Zeekr and Lynk & Co chase 1 million annual sales target after merger · TechNode NASA's new special Webb image shows chaos and creation Xiaomi to launch first electric SUV next spring, sources say · TechNode CATL seeks to manufacture batteries in the US pending Trump’s approval · TechNode Cybertruck recall: Tesla's dreaded, massive wiper is a big problem
1.9559s , 8223.96875 kb
Copyright © 2025 Powered by 【Oil Massage With Naked Body】,Inspiration Information Network