The complete professional position.
The Declaration on Responsible AI Use in Legal Practice, the Supervised Intelligence Method, the AI Responsibility Test, and the Five Doctrines of Professional AI Use. These instruments articulate no new obligations. They apply existing duties of diligence, verification, and independent judgment to current technological conditions.
Declaration on Responsible AI Use in Legal Practice
Part I sets out six principles governing AI-assisted legal practice. Part II, the Operational Framework Annex, contains the Supervised Intelligence Method, the AI Responsibility Test, the verification protocol, and a sample compliance clause.
System 1 Lawyering: The Cognitive Risk of AI-Assisted Legal Work
The white paper. Examines the cognitive mechanisms, including fluency-validity substitution, anchoring, coherence bias, and effort substitution, by which AI-generated legal content suppresses the independent verification that professional obligations demand. Each mechanism is mapped to documented judicial findings from Indian courts.
SIM One-Page Standard
The Supervised Intelligence Method in canonical form on a single page. Chamber-ready. The document you pin above your desk and apply to every matter.
Six principles.
Presumptive Status of AI Output
All AI-generated content shall be treated as presumptively unverified and unreliable until subjected to independent human verification against primary sources.
Non-Delegable Authorship Responsibility
The advocate bears complete and non-delegable authorship responsibility for every submission, argument, and citation filed or relied upon in any proceeding.
Mandatory Independent Verification
Verification of citations, statutory provisions, and legal analysis constitutes a core professional duty that cannot be outsourced to any technological system or platform.
Platform Independence
Professional standards governing the use of artificial intelligence apply uniformly across all tools and models and are not tied to the features or verification mechanisms of any specific platform.
Preservation of Professional Judgment
All strategic, doctrinal, and tactical judgments must remain the product of the advocate's independent analysis. AI systems may assist in research or drafting but cannot substitute for the advocate's cognitive engagement.
Disclosure and Structured Supervision
Where the use of artificial intelligence is material to the preparation of a submission, the advocate may disclose such use and confirm compliance with a structured supervisory protocol ensuring independent verification and judgment at every material stage.
The Supervised Intelligence Method.
Five stages governing every AI-assisted legal task. Three exclusively human. Two AI-assisted under continuous advocate direction. No stage may be inverted, skipped, or delegated.
Legal Framing
The advocate alone defines the legal question, identifies the applicable framework, and determines the issues. No AI participates. If the legal theory is chosen by AI, every stage that follows is built on a foundation the advocate did not lay.
Pattern Expansion
AI tools assist in broadening the search: surfacing authorities, analogous cases, procedural precedents, and legislative history beyond what the advocate would reach alone. Every output is presumptively unverified. This is where AI adds velocity without adding judgment.
Doctrinal Reconstruction
The advocate applies legal reasoning to candidate materials, reconstructs the argument, selects authorities, and determines how law applies to facts. Editing preserves the AI's structure. Reconstruction replaces it.
Verification
Every citation confirmed against primary sources. The actual judgment located and read. Every statutory provision confirmed against enacted text. Grounded tools may assist the search; responsibility for the verification outcome is solely the advocate's and may not be delegated, because no machine can answer for it before a disciplinary committee. The documented failures before Indian courts occurred at this stage.
Strategic Judgment
Which arguments to advance, which to withhold, how to sequence submissions before a bench, what tactical concessions to make. This requires knowledge of the court, the bench, and the client. The machine processes; the advocate authors.
The AI Responsibility Test.
Four questions an advocate must be able to answer before submitting any AI-assisted work, and before challenging the opponent's. The ART operates in both directions: inward on your own work product, outward as a forensic instrument against opposing submissions.
Framing Diligence
"Did I personally identify the legal issue before engaging AI?"
Supervisory Adequacy
"Did I maintain meaningful oversight of AI-generated content throughout?"
Verification Completeness
"Did I independently verify every citation, fact, and legal proposition?"
Judgment Independence
"Did I exercise independent professional judgment in final strategic decisions?"
Five Doctrines of Professional AI Use.
Cognitive Sovereignty
The advocate's independent judgment is non-delegable and sovereign. AI may assist, never substitute.
Epistemic Diligence
Every AI output is presumptively unreliable until independently verified. Partial diligence is professional negligence.
Fiduciary Confidentiality
Client confidences are held under absolute fiduciary duty. No client data enters any AI system without complete control.
Adversarial Anticipation
Assume opponents, corporations, regulators, and the State are already deploying AI at scale.
Intellectual Amplification
Governed by Doctrines I to IV, AI becomes a legitimate force multiplier expanding research and strengthening advocacy.
The Supervised Intelligence Method is not a recommendation. It is the operational standard.
AI for Indian Advocates · 2026The complete doctrinal treatment, with the constitutional and professional obligation framework, worked examples, and the adversarial toolkit, is in AI for Indian Advocates, Volume I. The research programme extending the standard is recorded on the Research page.
Field Notes, dispatched.
Practitioner briefs on AI, law, and professional responsibility, written from inside the courtroom. Sent occasionally. No noise.