Shivam Shukla
Canon · Policy and professional regulation

Refusal capacity: the activation and construction pathways

The capacity of an institution to refuse a harmful AI deployment on behalf of a profession or workforce and make the refusal bind. It arrives by two pathways that are one process at different stages: activation, where a statutory professional body already holds the power and has not used it; and construction, where no such body exists and a workforce organises, wins recognition and assumes the power itself.

A unit of the Library · Shivam Shukla, Advocate, High Court of Judicature at Allahabad
In practice

When this decides something for you

A regulator, a Bar Council member or an association officer is asked whether the body can say no: to a court-management system that summarises pleadings, to a firm's adoption of a drafting tool without a supervision rule, to an employer's use of a worker's likeness. The usual answer is that the body issues guidance and has no power to refuse. This page says the power usually exists, chartered in statute, and has not been exercised; and where it does not exist, the recent past shows how it is built.

The question

The question

Which institution, in a given profession or workforce, can refuse an AI deployment and make the refusal bind, and what does a profession do when that institution is asleep or absent?

The premise

The premise

A machine may verify, but a machine cannot be a respondent. Responsibility cannot be delegated to what cannot be held to account. So the refusal a profession needs has a principled criterion, answerability: a deployment is refused where it would leave no accountable person answerable for the work it produces. The institution that can apply that criterion and make it bind is the accountability layer, and most frameworks for governing AI lack it.

Mechanism

How it works

The activation pathway. In most established professions the institution already exists: a statutory self-governance body, chartered in public law, holding rule-making and disciplinary jurisdiction over every practitioner, and never switched on for this purpose. Activation is a court, a scandal or an organising drive turning it on. It requires no new statute and no new regulator; it reaches for the body that already holds the power. The contrast with the market is instructive: private infrastructure around professional work is being built at speed, while the public infrastructure of accountability over that work already exists and sits unused.

The construction pathway. Where no statutory body exists, the same institution is built from below. In 2023 two American screen guilds, with no statute behind them, struck and won bargained instruments: the Writers Guild of America's memorandum of agreement establishing that AI-generated material is not literary or source material, that no writer can be required to use AI, and that companies must disclose AI-generated inputs; and SAG-AFTRA's consent and compensation requirements for digital replicas of performers. In 2026 both floors survived renewal and were extended. A floor that holds through a full renegotiation cycle is a load-bearing structure, not an episode.

One process. The two pathways look like alternatives and are a sequence. Most professional councils were not handed down from above; they were won, by practitioners who organised as voluntary associations until the state wrote their collective power into statute. Today's dormant councils are the successful conclusions of earlier organising drives. A statutory professional body is a public-cooperative hybrid: chartered and accountable in public law, governed by its practitioner members, holding delegated power that neither a state agency nor a cooperative holds alone.

The map of refusal capacity across professions and jurisdictions, from a works council's statutory veto to a workforce with no body at all, is read not as a ranking but as a repertoire: every profession stands on an identifiable rung, and a documented mechanism exists for climbing. Its central empirical finding is that almost all professional-body responses to AI are soft: guidance that annotates existing duties rather than a rule that refuses.

The case

One case

Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd, 2026 INSC 668, Supreme Court of India, 2 July 2026. Having set aside tribunal decisions that rested on fabricated precedents, the Court declared zero tolerance for reliance on fabricated material at the Bar and on the Bench, held that "mere declaration of prohibitory action is not sufficient; there must be a consequential action following accountability", and directed the Bar Council of India, "being the apex statutory body, to constitute a committee and deliberate... and prescribe a guiding principle to prevent such occurrences, along with the disciplinary action that will follow a violation of the norms". The Court did not draft the rules itself and did not petition Parliament for a new regulator. It reached for the institution that already held rule-making and disciplinary power over every advocate in the country and had never been switched on for this purpose. That is activation, not construction. So far as the author has been able to document in English-language sources, it is the first instance of a court ordering a statutory professional regulator to govern AI; the author would be glad to learn of an earlier counterpart.

The rule

The rule

Refusal capacity is exercised or lost. A council that never says no becomes, in function, a council that cannot; a workforce that never organises remains one that may not. Where the body exists, activate it; where it does not, build it; the two are the same institution at different moments of its life.

Distinctions

Not to be confused with

Refusal capacity is not a veto on technology and not a demand for a new regulator. It is the power an existing or newly built body holds to refuse a specific deployment on the criterion of answerability, and to make that refusal bind on its members and, through them, on those who employ or instruct them.

ProvenanceSources · Method

Provenance and method.

Where this was published

  • Shivam Shukla
  • The Power to Say No: Activating the Missing Accountability Layer of the Solidarity Stack
  • SSRN 7454698 (2026)
  • sections 4
  • 5 and 8
AI Work RecordWritten from the author's published work (Shivam Shukla; The Power to Say No: Activating the Missing Accountability Layer of the Solidarity Stack; SSRN 7454698 (2026); sections 4; 5 and 8) under the Supervised Intelligence Model. Legal framing and the source text are the author's; drafting and the voice pass were machine-performed under the author's voice file; citations are confined to authorities the author has verified against archived primary records; published on the author's approval, 2026-09-25; version 1.

Changelog

  • 2026-09-25, version 1. First published.

Read next in the Library