The 2026 C:\>DIR Global ‘Agentic Regulator’ Hackathon Report – Process, Lessons Learned & Recommendations examines a different approach to developing technology for public authorities: giving regulators a direct role in building and testing solutions to the problems they face. Its co-authors are affiliated with Financial Innovation for Impact (Fii), the Cambridge Centre for Alternative Finance (CCAF) at Cambridge Judge Business School, the Global Finance & Technology Network (GFTN), and Ofcom.
The 2026 Global AI in Financial Services Report found that 40% of surveyed financial firms had reached advanced stages of AI adoption, compared with 20% of regulators. Agentic systems can select and carry out steps towards a goal using connected tools. Their use raises questions across financial regulation, consumer protection, competition and data governance, often involving several authorities.
Building tools within a week
The Hackathon attracted 942 applicants from more than 65 countries. Nineteen of the 36 finalist teams came from, or were affiliated with, public authorities. Working with accessible AI models, agent frameworks and synthetic datasets, teams developed prototypes during a one-week build period. Policymakers and supervisors contributed their knowledge of regulatory workflows and the decisions those workflows support.
The overall winner, Team Pheonix from the Securities and Exchange Board of India, used trading agents in a simulated securities market to test proposed rules before implementation.
Work covered six problem spaces:
Agentic Financial and High-Stakes AI Advice
AI-Driven Fraud and Scams
Agentic Payments, Commerce and Their Oversight
Know Your Agent (KY-A), Digital Verification and Digital Public Infrastructure
Decentralised Market Infrastructure, Smart Contracts and AI Agents
Market Manipulation, Agentic Herding and Stability
Four of the six winning teams were from, or affiliated with, emerging markets and developing economies.
The Hackathon was supported by 48 global partners. The wider network included the Bank for International Settlements Innovation Hub (BISIH), GFTN, Digital Regulation Cooperation Forum (DRCF), NFH, Ethereum Foundation, Swift, Google, Moneybox, Ant International, Euroclear, and many other organisations. Moneybox, Ant International and the Ethereum Foundation also supported individual problem spaces; NayaOne provided the common build environment; and vouch.finance provided seeded supervisory environments.
From prototypes to public use
The report examines the risks teams identified and the safeguards they implemented or proposed. The safeguards fall under four themes: human-in-the-loop, auditability and traceability, cyber risk management, and safety and governance. Examples include limiting access to tools and data, preserving records of agent activity, and requiring human approval for consequential decisions.
The demonstrations showed what could be built in a short period. Further testing is needed to establish whether the applications are reliable in everyday use. Connecting these applications to authorities’ existing data and systems, and obtaining governance approval, also remain substantial tasks.
The recommendations consider how authorities could support this work. Early applications could focus on narrow, verifiable tasks, using synthetic or closed-case data. Authorities could also assess whether existing commercial products can be adapted to their needs, with arrangements to protect sensitive supervisory information and address conflicts of interest.
Shared development and joint procurement could help authorities meet common needs, while allowing each institution to adapt tools to its mandate. The report also recommends investment in training and an innovation culture in which senior management give staff time and support to experiment.
Teams also highlighted the need for support to take their prototypes into use. An incubator or accelerator could provide this support, for example through regional development hubs. These could offer a setting to test solutions with real-world data and run pilots with selected authorities, supported by common governance guidance. Tested solutions could then be adapted and deployed by other authorities at low cost.
“Regulators cannot oversee an agentic financial system with yesterday’s tools, and this hackathon showed they don’t have to. Regulators can become builders. Nineteen of our 36 finalist teams came from central banks and regulatory authorities, and they vibe-coded working prototypes in a week, because nobody knows a supervisory workflow better than a supervisor. The private sector brought great solutions too, and the way forward is for the public and private sectors to work closely together to create agentic solutions for regulators, fast and at scale.”
Bryan Zhang, Co-Founder and Executive Chair, Financial Innovation for Impact, and Executive Director, Cambridge Centre for Alternative Finance
Read the findings and recommendations from the C:\>DIR Global ‘Agentic Regulator’ Hackathon, including practical examples of how public authorities are developing and testing agentic AI.
Explore the report, 2026 C:\>DIR Global ‘Agentic Regulator’ Hackathon Report – Process, Lessons Learned & Recommendations.