Today, onn August 11th, a coordinated effort is taking place in Las Vegas hotels as a federal group of hackers will be trying to penetrate some of the world’s most potent artificial intelligence systems for three consecutive days. This “red-teaming” assault, initiated by the Biden administration and set to coincide with a hacker conference, aims to uncover vulnerabilities in these AI systems. Industry leaders such as OpenAI, Google, and Meta agreed to subject their cutting-edge AI models to this exercise. The endeavor underscores a primary concern of the White House: managing the risks associated with rapidly advancing AI models is necessary to harness their potential benefits. Particularly with extensive language models, these risks are considered to be multifaceted, less explicit than other systems, and including various aspects of society, economy, and national security.
Of course these concerns are correct. And what about the future government – run generative AI models? Who will protect us from our protectors?
