2 maiAffirmative Safety: An Approach to Risk Management for Advanced AiThe paper suggests that companies developing high-risk AI systems should demonstrate their safety before deployment, arguing for...