How AI Is Reshaping International Arbitration: Key Insights from William Kirtley's Istanbul Talk
At a conference in Istanbul marking the merger of Özcan Legal and Fırat Gültekin & Partners, William Kirtley laid out a clear view of where artificial intelligence is pushing international arbitration. The message was direct: AI is already changing advocacy, procedure, and decision-making, and the pace is accelerating.
For counsel and arbitrators, this is no longer theory. It is day-to-day practice. The practical question is how to adopt AI without compromising independence, confidentiality, or due process.
Where AI Already Delivers Real Value
- Case preparation and drafting: Tools now produce clean first drafts that resemble junior-associate output. They reduce typos and inconsistent language while speeding up iterations.
- Research and analysis: Faster answers, better triangulation of sources, and automated cross-references. Still requires verification, but it shortens the path to usable arguments.
- Document review: Efficient screening of large productions, prioritizing relevant evidence, and summarizing technical materials in complex disputes.
- Strategy inputs: Early case-assessment, argument testing, and data-informed arbitrator selection are increasingly common inside teams.
How Tribunals Are Using AI
Tribunals are starting to lean on AI for administrative tasks, summarizing submissions, and drafting preliminary reasoning. The aim is to cut noise and focus time where judgment truly matters.
The most notable development is the AAA/ICDR pilot introducing an AI-powered arbitrator for lower-value, document-only cases. A human arbitrator still reviews and signs the award, but the pilot raises hard questions about transparency, accountability, and how far delegation can go.
Legal and Ethical Fault Lines
- Hallucinations: AI can generate confident but wrong statements. Citations and quotes must be checked every time.
- Bias: Training data and prompt design can skew outputs. Parties should anticipate challenges to fairness if models or settings go undisclosed.
- Confidentiality: Cloud tools may create exposure under protective orders or data-transfer rules. Contract terms and technical controls matter.
- Discoverability: Prompts, system messages, and drafts may be sought in disclosure. Treat prompts like work product and manage them accordingly.
These issues are no longer hypothetical. In LaPaglia v. Valve Corp., a party seeks to vacate an award alleging the arbitrator relied too heavily on AI. The case is unresolved, but it shows how parties and courts will test the boundaries of acceptable use.
Soft-Law Guidance Is Converging
Recent guidance from institutions and professional bodies points in the same direction: decision-making stays with humans, oversight is essential, and disclosure may be required where AI use could affect fairness. Expect updates as tools improve and case law develops.
What Counsel and Tribunals Can Do Now
- Adopt an AI use policy: Define permitted tools, confidentiality rules, verification steps, and approval workflows.
- Disclose thoughtfully: Where AI materially affects submissions or procedure, consider narrow, purposeful disclosure that protects privilege and strategy.
- Keep humans in the loop: Require attorney sign-off on all AI-assisted drafts, fact summaries, and legal analysis.
- Verify citations and quotes: Cross-check sources, dates, and pin cites. No exceptions.
- Secure data: Use enterprise accounts, encryption, and opt-outs from data retention where available. Avoid pasting sensitive data into public tools.
- Maintain prompt hygiene: Store prompts securely, redact client identifiers, and keep versioned logs to explain decisions if challenged.
- Be explicit on arbitrator selection: Treat analytics as inputs, not instructions. Document reasoning beyond model outputs.
- Train juniors differently: Shift from first-draft writing to reviewing, testing arguments, and fact development. Teach them to supervise machines, not copy them.
The Role of the "AI Conductor"
Kirtley's closing point was simple: lawyers will spend less time drafting from scratch and more time directing, verifying, and refining machine output. Judgment, ethics, and strategic clarity become the differentiators.
Digital literacy is now a baseline skill. For legal teams building capability, curated courses can shorten the learning curve without sacrificing standards. See a practical selection by role at Complete AI Training.
Filed Under: Artificial Intelligence Arbitration, William Kirtley
Your membership also unlocks:
