Read news on theoretical concept with our app.
Read more in the app
As AI models start exhibiting bad behavior, it’s time to start thinking harder about AI safety | AIs that can scheme and persuade were once a theoretical concept. Not anymore.