Responsible Innovation

Responsible innovation is usually framed as a constraint on invention — a brake, applied by people who do not build things, to slow down people who do. We think that framing is both wrong and unhelpful, and that the real variable is timing rather than restraint.

The cost of changing a decision rises steeply with time. At specification, a requirement is a sentence. At procurement, it is a contract clause. At integration, it is a change request. After deployment, it is a withdrawal, a remediation programme, a regulatory response and possibly litigation. Ethical review almost always arrives at the last of these points, by which time the contract is signed, the integration is partly built, the staff have been reassigned and the alternatives have been discarded. Review at that stage can document concerns; it cannot realistically stop anything, and everyone in the room knows it. The result is a process that is expensive, adversarial and ineffective — and which then gets cited as evidence that ethical review obstructs delivery.

We work upstream, where a decision can still be reversed cheaply: in requirements, in procurement, in standards and in design review. Concretely, that means clauses in tender documents rather than principles in strategy documents — audit access with a genuinely independent auditor, performance disclosure in the deployed population, override preservation with the override rate reported, exit terms agreed before dependency exists, and notice before material change. None of these is exotic, and none is expensive at tender. We are candid about the obstacle we cannot draft around: procurement teams are measured on cost and delivery time, and every clause above lengthens both. Until an institution measures its procurement function on the durability of what it buys, good clauses will remain available and unused.