XAI for Safety and Security: A Bottom-Up Approach

Despite there being many compelling examples of artificial intelligence (AI) being used for good, widespread public acceptance of it has been hindered by instances of decidedly harmful misuse. Widespread private development has likewise been hindered, saliently within the safety and security industry, by the fact that AI may behave in unexpected and inexplicable ways, sometimes exhibiting harmful bias. Companies are fulfilling their capacity to develop AI due to the consequences of unwitting breaches of legal frameworks, like GDPR and the AI Act. These regulations are designed top-down, from a regulatory perspective. They often lack a comprehensive technical perspective. Companies without resources to devote to legal compliance during development first develop their technical solutions, and subsequently check for legal compliance. With this project, we suggest dealing with legal and ethical aspects of AI solutions using a bottom-up approach.