Ethics Boards Without Teeth
An ethics board that cannot stop a deployment is not a governance mechanism; it is a press release with a seating plan.
The Season of the Ethics Board
Something has shifted in the past eighteen months. Organisations that had never previously expressed a public position on the ethics of their technology operations are now establishing ethics boards, ethics committees, and ethics advisory panels with conspicuous enthusiasm. The pattern is visible across sectors — financial services, technology, healthcare, public sector — and the timing is not coincidental. A combination of high-profile algorithmic failures, growing public scepticism about data practices, and the regulatory momentum established by GDPR has made ethics governance fashionable in a way it was not two years ago.
The question worth asking is whether these bodies are governance mechanisms or performance. In my observation, the answer is predominantly the latter.
The Authority Problem
The defining characteristic of most ethics boards I have encountered — directly or through the accounts of colleagues working within them — is the absence of consequential authority. They advise. They review. They publish principles. They convene quarterly, hear presentations from product teams, and produce recommendations that are forwarded to senior leadership for consideration.
What they cannot do, in almost every case, is stop anything. They cannot halt a product launch. They cannot mandate a redesign. They cannot require that an algorithmic system be withdrawn from operation pending review. Their authority is advisory, and advisory authority in a commercial environment where deployment timelines are set and revenue targets are committed is, functionally, no authority at all.
An ethics board that cannot stop a deployment is not a governance mechanism; it is a press release with a seating plan.
This is not a criticism of the individuals who serve on these boards, many of whom are serious scholars and practitioners with genuine expertise. It is a criticism of the organisational design that constrains them. The board is constituted to provide the appearance of ethical oversight without the operational integration that would make oversight effective.
The Independence Illusion
The second structural weakness is independence — or rather, its absence. Most ethics boards are appointed by the organisations they oversee, funded by those organisations, and provided with information curated by those organisations. Their terms of reference are set by the organisation. Their access to systems, data, and internal deliberations is mediated by the organisation.
This is not independence. It is consultation on terms defined by the party being consulted about. The parallel with audit is instructive: the accounting profession spent decades establishing the principle that an auditor must be independent of the entity being audited, precisely because the value of the function depends on the auditor’s ability to reach conclusions that the entity may not welcome. Ethics boards, as currently constituted, have not learned this lesson.
The organisations that are serious about ethical governance — and there are some, though fewer than the public announcements suggest — grant their ethics functions genuine independence: separate reporting lines, independent access to systems and data, the authority to commission their own reviews, and the power to publish findings without prior approval.
What Would Real Governance Look Like
Real ethics governance would be embedded in the product development lifecycle, not appended to it. It would operate at the point where decisions are made — during design, during data selection, during model training, during deployment planning — not after the fact in a quarterly review meeting.
It would have three characteristics that most current arrangements lack:
- Decision rights. The ethics function would have the authority to require changes to — or halt — initiatives that fail to meet defined ethical standards, with escalation to the board only where the ethics function and the business cannot agree.
- Operational access. The ethics function would have direct, unmediated access to the systems, data, models, and decision logs it needs to conduct meaningful review — not a curated summary prepared by the team under review.
- Accountability symmetry. The consequences of ignoring ethical guidance would be as concrete as the consequences of missing a delivery date — reflected in performance assessments, in project governance, and in the risk register.
None of this requires new technology or new regulation. It requires organisational will — the willingness to accept that ethical oversight, like financial audit, must sometimes produce conclusions that the organisation finds uncomfortable, and that the value of the function depends entirely on its power to do so.
The current generation of ethics boards, well-intentioned as many of them are, does not meet this standard. Until it does, the governance they provide will remain symbolic — reassuring to the public, useful for recruitment marketing, and irrelevant to the decisions that actually matter.