Outcome standard and method standard
Two kinds of professional standard kept apart. An outcome standard states what finished work must not contain (a pleading must not cite an authority that does not exist). A method standard states how the work must be done for the outcome to be reliable (what checking, what supervision, what record constitutes adequate control). Generative AI changed neither outcome standard; it created a method that no method standard yet governs.
When this decides something for you
A practitioner is told that AI changes nothing, because the duty to verify always existed. A regulator is told the same, and issues a note saying that existing duties apply. Both statements are true and both miss the question the practitioner actually has: whether checking a machine's citation against a second machine discharges the duty; whether a junior's use of a chatbot preserves the chain of supervision; what record of checking a tribunal will later accept. Those are questions about method. Seeing that they are, and that the answer to them has not been written, is what this page is for.
The question
What did generative AI change in the standards that govern professional work, and what did it leave untouched?
The premise
Every mature profession runs on two kinds of standard. Outcome standards say what the finished work must not contain: a pleading must not cite authorities that do not exist; an audit must not certify accounts that misstate; a diagnosis must not ignore the presenting symptom. Method standards say how the work must be done to make those outcomes reliable: what checking, what supervision, what independence of judgment constitutes adequate professional control.
How it works
Outcome standards are ancient and unchanged. The prohibition on citing fabricated authority is as old as citation; no advocate in any legal system was ever free to invent a precedent. What generative AI changed is the economics of breaching the standard. Fabrication that once required deliberate fraud now requires only ordinary negligence at scale, because the machine produces fluent, formatted, confident falsehood as a by-product of normal operation. The same outcome standard is now breached by people who intended no breach.
Method standards are written when a new method arrives, and this one has arrived without its standard. A rule that says "do not file fabricated authority" tells the practitioner nothing about whether machine cross-checking satisfies the duty of verification, whether delegation to a junior who used a chatbot preserves or breaks supervision, or what record of checking will count as diligence. Those are method questions. Where no body has answered them, they are answered by whoever sanctions the practitioner, after the event, in the terms of that case.
The distinction disposes of two familiar objections. To "the duty always existed": the outcome duty did; the method standard did not, because the method it must govern did not exist. To "guidance is everywhere": guidance that says existing duties apply restates the outcome standard in a new context; it does not state the method. Mapping is not promulgation.
The distinction also says what a method standard for this method must contain. Not "be careful", which is an outcome standard in different words, but: which stages of the work the machine may enter; what the human must do at each; what is verified against which class of source; what record is kept and for how long. A staged method that reserves framing, verification and judgment to the human, with a four-step test of whether that was done, is one such standard; any body with the authority to bind can adopt one like it.
One case
Gummadi Usha Rani v. Sure Mallikarjuna Rao, SLP(C) No.7575 of 2026, Supreme Court of India, order dated 27 February 2026. Four non-existent decisions in a trial court's order breached an outcome standard as old as law reporting. No one contends the trial judge intended to invent them. The Supreme Court declared that a decision so made "is not an error in the decision making. It would be a misconduct and legal consequence shall follow", and issued notice to the Bar Council of India. Read with the distinction on this page, the order is the outcome standard being enforced against a breach that the absence of a method standard made ordinary, and the notice to the regulator is the Court pointing at where the method standard has to come from.
The rule
The outcome standard is old and binds today. The method standard for generative AI is unwritten, and until the body with the power to write it does so, it is being written one sanction at a time.
Not to be confused with
A method standard is not a technology standard. It does not certify tools, set accuracy thresholds or approve vendors. It governs what the professional does with any tool, and it is written by the profession's own body, not by the tool's maker.
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)
- section 2
Changelog
- 2026-09-25, version 1. First published.