The proliferation of AI ethics frameworks, principles documents, governance guidelines, ethical codes, is itself a phenomenon that postmodern analysis can illuminate. Rather than treating these frameworks as straightforward expressions of ethical commitment, postmodern analysis asks: what assumptions do they embed? What do they include and exclude? Whose interests do they serve?

A deconstructive reading of the major AI ethics frameworks reveals several recurring features.

The liberal individual as default subject. Most AI ethics frameworks centre their concern on the individual, individual rights, individual autonomy, individual fairness. The structural dimensions of AI harm, the ways in which AI systems produce and reproduce collective inequalities, receive less attention. This reflects the liberal philosophical tradition from which most ethics frameworks derive, in which social problems are primarily problems of individual rights and their protection.

Technical solutionism. The predominant mode of ethical AI frameworks is technical: identify the problem (bias, opacity, inaccuracy), develop a technical solution (debiasing algorithms, explainability tools, accuracy metrics), implement and evaluate. This framing positions ethics as a technical problem that technical experts are best positioned to solve, and marginalises the political, economic, and structural questions that CRT, feminist, and Marxist theory insist are central.

The universality claim. Many AI ethics frameworks claim to articulate universal principles, fairness, transparency, accountability, that apply across cultures, contexts, and social positions. Postmodern analysis is sceptical of such claims. Universal principles tend to reflect the values of the most powerful groups, those with the resources and authority to participate in standard-setting processes. The OECD AI Principles, for example, were developed by the governments of wealthy countries and reflect, in many ways, the values and interests of those countries.

What the frameworks do not say. Deconstructive analysis attends as much to absences as to presences. Major AI ethics frameworks rarely address: the ownership structures of AI companies, the labour conditions of workers in the AI production chain, the concentration of AI capability in a small number of corporations, the political economy of AI regulation, or the historical and structural roots of the inequalities that AI systems reproduce. These absences are not accidental. They reflect the interests of the actors who participate in framework development.

This analysis does not imply that AI ethics frameworks are worthless. It implies that they should be read critically, attentively to what they do and do not say, and with awareness of whose interests shape their content.

Reflection question: Read the AI ethics principles of an organisation you know. What do they say? What do they not say? Whose interests do they primarily serve?