Should certain AI activities, uses or goals, be prohibited? or should AI be open to any field? Should AI technologies be classified according to the risks they involve? If so, what risks should be considered? Should we set up a pattern to define the level of AI risk? From then, how to proceed to mitigate the risks? Would it be effective to tackle the ethical challenges?<\/p>\n
In Spring 2021, the European Commission introduced a Bill, \u201cProposal for a regulation laying down harmonized rules on artificial intelligence\u201d which encloses a chapter on prohibited practices. If adopted, the proposal would implement an AI-risk management system, where AI practices would be assessed by their level of risks and comply with different obligations accordingly. In the meantime, amongst a multitude of bills introduced, the US Congress has adopted two Acts, first the \u201cAI Government Act 2020\u201d , and second \u201cThe National Artificial Intelligence Act 2020\u201d . However, under a direction from the Congress, the NIST is committed to define a \u201cconsensus-driven\u201d, \u201cvoluntary\u201d and \u201cguidance document\u201d also based on a risk management system AI Risk Management Framework (see ai.gov).<\/p>\n
Therefore, the workshop suggests having a general presentation and discussing those European and American approaches on public, private and broad perspectives, the current criteria set for defining AI levels of risk, the tools to assess them and the involved obligations, and studying how these patterns could, if adopted, \u201ccapture\u201d the ethical management of AI and deal effectively with AI ethics.<\/p>\n