The manifesto
A profession makes its commitments in public; accountability is the point. These are ours: the position on AI in legal practice, declared the way the word profess demands, out loud. Twelve of them, like the tables on which Rome first wrote its law.
To profess is to declare publicly and be bound. A professional is not someone who is paid to do something; a professional is someone who has professed, and remains answerable to that declaration every working day.
AI can produce competent legal work at machine speed and near zero cost, and we embrace it. But it has taken no vow, sits on no roll, owes no duty to the court and stakes nothing when it is wrong. What it produces is material, not advice; advice comes into existence at the moment a person who has professed takes responsibility for it, and that moment is the product, the protection and the point.
Intake, drafting, assembly, checking, research triage: the machine should do the first pass of all of it, and clients should feel the benefit in speed and honest fixed fees. Refusing the machine is not professionalism; it is nostalgia billed by the hour.
Clients come to lawyers for experience, expertise and judgement calls; for qualifications and certifications; for rules, principles and ethics. Where these are sought, a human carries the work, visibly and accountably.
Its outputs are not revelations to be received on faith, however fluent the prose. AI is a product of humanity and of nature, trained on our words, our judgments and our mistakes; it inherits our imperfection at scale and at speed, wrapped in a fluency that hides it. Machine output deserves the treatment we give any persuasive witness: verification. That is not hostility to the technology; it is the technology taken seriously enough to check.
In other domains the human in the loop is a checkpoint, and a review with a button click will do. In law the human is the product. A profession is a deliberate narrowing of who may advise and how quickly when someone's life or business depends on it; we built the narrow place on purpose, and we put a person in it.
A lawyer who clicks approve on machine output they have not genuinely examined has not supervised it; they have counterfeited the signature. AI errors in law are supervision failures, not technology failures.
Courts, tribunals and parties exist in conflict about what the law is, what the evidence proves and whose interpretation should prevail. The system admits as much itself: every right of appeal is the law accepting its first answer may be wrong, and every dissenting judgment preserves the argument just rejected, in case the future needs it. One immutable and definite answer never truly exists, and where something close to one does, it holds only until a higher court, a new statute or a better argument moves it. The machine can state the consensus; it cannot sit inside the disagreement, and the disagreement is where law is practised.
A law can be validly passed, on the books and enforced by the courts, and still be wrong. Apartheid was law; slavery was law; colonialism was administered through law. The capacity to apply the law while refusing to mistake it for justice is not a feature that can be trained in; it is a conscience, and it comes only with the person.
Confidentiality, privilege and the duty of care exist because someone professed. No client file goes into unsupervised public AI tools, because casual AI use can destroy the very protections professing creates.
Every AI-assisted matter should carry an audit record of what was checked, changed and rejected, ending in a named professional's signature. Accountability that cannot be evidenced is sentiment.
The machine may do everything up to the line. The line itself belongs to the professed, and it is the one thing we will not let the machine touch.
The machine may do everything up to the line. The line itself belongs to the professed.
Chris Dias
The manifesto is the position; the thesis is the reasoning, from the Latin profiteri to what the law has always known about itself.
Read the thesis