A New Method to Steer AI Output Uncovers Vulnerabilities and Potential Improvements
February 19, 2026
February 19, 2026
LA JOLLA, California, Feb. 19 -- The University of California San Diego campus posted the following news:
* * *
A New Method to Steer AI Output Uncovers Vulnerabilities and Potential Improvements
*
Using this steering approach, the research team conducted experiments on some of the largest open-source LLMs in use today, such as Llama and Deepseek, identifying and influencing 512 concepts within five classes, ranging from fears, to moods, to loc . . .
* * *
A New Method to Steer AI Output Uncovers Vulnerabilities and Potential Improvements
*
Using this steering approach, the research team conducted experiments on some of the largest open-source LLMs in use today, such as Llama and Deepseek, identifying and influencing 512 concepts within five classes, ranging from fears, to moods, to loc . . .