Ethics as a service: a pragmatic operationalisation of AI Ethics

The "ethics washing" problem is real: The paper finds that existing tools designed to help developers implement AI ethics are either too flexible (allowing companies to claim ethical compliance without meaningful action) or too rigid (unable to adapt to different contexts), creating a significant gap between ethical principles and actual practice.

Principle-based codes alone are insufficient: While awareness of AI ethics issues has grown and led to numerous ethics guidelines and frameworks, the paper demonstrates that current legislation and regulations provide inadequate protection against AI harms to individuals, groups, society, and the environment.

A new approach is needed beyond technical tools: Rather than relying solely on technical guidance and translational tools to convert principles into practice, the paper suggests exploring alternative methods to make AI ethics genuinely useful for practitioners—implying that the traditional approach of principle-to-practice translation has fundamental limitations.

Context matters for ethical implementation: The paper identifies that one key failure of existing tools is their inability to be responsive to context, suggesting that effective AI ethics requires flexibility and situational judgment rather than one-size-fits-all solutions.

Operationalizing ethics requires rethinking the entire approach: The paper frames AI ethics as a service rather than just a set of principles or technical requirements, suggesting that organizations need to fundamentally restructure how they integrate ethical considerations into AI development processes.