AI companies have to commit to adopting a state-of-the-art safety and security framework to outline their systemic risk management processes, the EU’s Code of Practice for General Purpose AI says.