Using AI in Arbitration Responsibly
AI is already widely used with legal work, including arbitration. But standards around its use have not kept pace.
Who is responsible when AI gets something wrong? What should it be used for? When should its use be disclosed? And which decisions should remain entirely human?
Rule 38 of the CORD Rules, and CORD's AI Practice Note, provide a framework to answer these questions.
CORD's approach is not to prohibit AI. It is to allow AI's use, while making clear that the person using it remains responsible for how it is used, and for the work it produces.
Clear rules for different uses
Parties and their representatives may generally use AI without disclosure, but remain fully responsible for everything they submit.
Witnesses and experts may use AI to organise and present information, but must disclose its use where AI performs work that would ordinarily be expected to be their own.
For arbitrators, the AI Practice Note goes further. It identifies uses that require no disclosure, uses that require prior disclosure, and uses that are prohibited altogether.
- Organising documents
- Preparing chronologies
- Making calculations
- Checking citations
- Producing convenience translations
- Legal research
- Analysing evidence
- Summarising submissions
- Preparing procedural orders
- Drafting any part of an award on the merits
- Deciding issues of fact or law
- Assessing witness credibility or weight of evidence
- Determining damages
- Delegating any decision-making function
AI assists. Humans remain responsible.
Rule 38 brings accountability to the centre of arbitration. Parties and their representatives remain responsible for the veracity and accuracy of their AI use, and arbitrators remain responsible for work undertaken with AI assistance.
The Rules also protect confidentiality. Anyone using an AI tool must ensure, among other things, that confidential arbitration data is not improperly stored and that the tool does not train on the arbitration data.
Why it matters
The question is no longer simply whether AI should be used in arbitration. It is how it should be used responsibly.
Rule 38 gives parties, counsel, and arbitrators clear boundaries. AI may assist with an arbitration, but responsibility for AI use — as well as the responsibility for deciding a dispute — remains wholly and exclusively human.