Ethics washing

Ethics washing is the framing that names the deployment of AI ethics discourse — principles documents, advisory boards, high-level expert groups, and self-regulatory commitments — as a public relations and lobbying instrument that forecloses binding regulation without producing substantive accountability. The framing's central claim is that industry actors who publicly embrace "ethical AI" are doing so not to accept constraint but to occupy the regulatory space that binding law would otherwise fill: to appear to have answered the accountability question while keeping enforcement off the table. It is a meta-critique — a critique of the form of accountability being offered, not only of specific failures within it — and has become a standard vocabulary item across AI accountability civil society on both sides of the Atlantic.

Origin

The framing's specific entry into the AI domain came through the European AI ethics guidelines process of 2018–2019. The EU Commission had convened a 52-member High-Level Expert Group on Artificial Intelligence to develop guidelines for trustworthy AI; the composition gave industry representatives roughly half the seats. When the "Ethics Guidelines for Trustworthy AI" were published in April 2019, philosopher Thomas Metzinger — one of the Group's members and a participant in drafting — published a sharp critique in Der Tagesspiegel, republished by the Business & Human Rights Resource Centre, calling the result "ethics washing made in Europe." Metzinger's characterisation was specific: he had watched industry-affiliated members successfully remove non-negotiable red-line language from successive drafts. AlgorithmWatch and Netzpolitik.org's contemporaneous investigation documented what had been deleted — prohibitions or strict limits on autonomous weapons systems, citizen scoring, and mass facial-recognition surveillance — establishing the concrete record beneath the abstract critique. Metzinger's characterisation in that investigation was the capsule formulation that carried: companies, he argued, "engage in an ethical debate with the goal of postponing or completely preventing legal regulations."

The framing's intellectual consolidation in the English-language register came in December 2019 with Rodrigo Ochigame's "The Invention of 'Ethical AI': How Big Tech Manipulates Academia to Avoid Regulation" in The Intercept. Ochigame argued that Silicon Valley's promotion of AI ethics was not a genuine ethical turn but a lobbying strategy — documenting corporate funding flows into academic AI-ethics initiatives including the Fairness, Accountability, and Transparency (FAccT) conference and university-based centres — and naming the phenomenon for the wide readership The Intercept commands.

Propagation

United States — from critique to institutional alternative

The same month as Metzinger's op-ed, April 2019, a concrete instance of the framing's subject arrived: Google announced its Advanced Technology External Advisory Council (ATEAC) — an AI ethics advisory board — and then dissolved it less than two weeks later following protest by over 2,300 Google employees and more than 300 civil-society signatories. Critics noted that the board's structure — an unpaid advisory function with no enforcement authority and four scheduled meetings per year — illustrated precisely the structural limitations the ethics-washing critique named: the form of an accountability mechanism without the substance of one.

The December 2020 firing of Timnit Gebru from Google's Ethical AI team — after Google retracted a paper she co-authored on the risks of large language models — became the most widely cited instance of the ethics-washing dynamic in practice, and crystallised a response: Gebru founded the Distributed AI Research Institute (DAIR) in December 2021, explicitly outside corporate and government funding structures, as an institutional form designed to make the independence that ethics-board review could not. DAIR's critique — that "AI ethics" at major labs functioned as institutional cover, and that genuine accountability required independence rather than internal review — operationalised the ethics-washing framing as an organisational practice.

Europe — regulatory campaigning

In Europe, the ethics-washing framing moved from academic critique into the live regulatory debate over the EU AI Act. AlgorithmWatch, the European Digital Rights network (EDRi), and Access Now coordinated multiple waves of civil society statements through 2021–2023 — ultimately signed by more than 150 organisations — demanding binding rights protections rather than the voluntary ethics-compliance and risk-management mechanisms the Commission's initial draft privileged. The framing's practical edge in this context was its diagnostic function: advocates could point to the EU's own 2019 guidelines experience as the demonstrated failure mode, and argue that the AI Act's drafting under industry lobbying pressure risked reproducing it.

Academic consolidation (2021–2022)

By 2021–2022, the framing was being systematised in peer-reviewed literature. Ben Green's 2021 "The Contestation of Tech Ethics" (Journal of Social Computing) identified "ethics washing" as a corporate strategy operating through the capture of ethics discourse — embracing the language of ethics to defuse criticism and resist government regulation, without committing to ethical behavior. Gijs van Maanen's 2022 "AI Ethics, Ethics Washing, and the Need to Politicize Data Ethics" in Digital Society offered the fullest academic consolidation of the critique: arguing that dominant academic ethics frameworks enable corporate "ethics shopping" for principles that minimally constrain behaviour, and that the framing requires a more explicitly political data-ethics response — one grounded in activist work rather than principle-only frameworks.

Why it has carried

Three features explain the framing's reach across the European regulatory debate, the American institutional-critique tradition, and the scholarly AI accountability literature.

First, the framing analogises to a term the public already knows. "Ethics washing" follows the pattern of "greenwashing" — the accusation that environmental commitments are deployed as cover rather than constraint — giving advocates an immediately comprehensible frame that pre-loads the critique. The analogy also carries a practical implication: just as greenwashing generated pressure for binding environmental disclosure requirements, ethics washing can be used to argue for binding AI accountability requirements.

Second, the framing is diagnostic rather than accusatory in form. It names a structural dynamic — the substitution of voluntary gestures for enforceable rules — rather than attributing bad intent to individual actors. This makes it usable across coalitions that may disagree about which companies are behaving in good faith: the claim is structural, and even genuine good-faith ethics commitments can be tested against it (whether they change behaviour that would otherwise occur, and whether they substitute for or complement binding requirements).

Third, the framing is validated continuously by real-world events. The Google ATEAC dissolution (April 2019), the Gebru firing (December 2020), and the documented weakening of successive EU policy documents under industry lobbying pressure each serve as cited instances — not hypothetical risks but observable cases that fit the described pattern. The accumulation of these cases is itself part of the framing's propagation; each new instance is legible through the vocabulary the framing established.