How banks can stop scams together
Explore the pilot and evidence ↓
ZQUAS is taking part in the FCA's Supercharged Sandbox. The FCA does not endorse ZQUAS, its products or its results.
Join the pilot
TRL 6. Synthetic data only. No production deployments.
Banks and payment firms
Check the payments coming into your own accounts, on your own data, inside your own installation.
Join the pilotCrypto exchanges
Exchange participation is being retested with information available to each exchange at the decision time.
Join the pilotInvestors and partners
A governance engine with measured results on synthetic data. Financial crime is where it starts.
Talk to usStudies on synthetic data
An offline retrospective study identified payments for review in one synthetic world. It did not measure executed prevention under a live review queue.
Exchange participation is being retested with information available to each exchange at the decision time.
The reported timing measures the warm decision path, not cold arrival-to-decision delay.
Read the test detail
Up to 540 simulated bank installations ran together on nine cloud GPUs in one cloud zone.
The prototype supports bounded information sharing and signed decision records under the stated trust model. Synthetic data. TRL 6. No production deployments or live counterparty data.
Built on a governance engine, not a fraud model.
The prototype supports bounded information sharing and signed decision records under the stated trust model. Financial crime is where we start. The same engine can govern decisions in other industries where automated decisions need rules, limits and an audit trail.
The platform →Explore the details
How the network works
Bounded information sharing and signed decision records.
The receiving bank checks its own accounts
The bank that holds the receiving account checks the payments coming into it. It uses its own view of that account and confirmed scam reports shared across the network. When an account looks like a mule, that bank can review it before more money arrives, even when the victim's bank is not in the network. A paying bank can also ask the holding bank one narrow question about a new payee before the money leaves.
Bounded sharing of confirmed reports
The prototype supports bounded information sharing and signed decision records under the stated trust model.
Bounded information sharing
The prototype supports bounded information sharing and signed decision records under the stated trust model.
The scam problem and its sources
Scams cost UK victims £576.4 million in 2025.
Authorised push payment (APP) scams trick victims into sending the money themselves. The money goes to a mule account, an account the criminals control. One mule account often collects from victims at several different banks, and the money is quickly moved on. Each bank sees only its own customer and its own payment.
Lost by victims to APP scams in the UK in 2025, up 19%.
APP scam cases in the UK in 2025.
Reimbursed by banks in 2025, 61% of losses.
Source: UK Finance, Annual Fraud Report 2026 (figures for 2025).
Under the UK's mandatory reimbursement rules, in force since October 2024, the sending bank and the receiving bank split eligible reimbursement under the applicable UK APP rules 50:50, up to £85,000 a claim.
Source: Payment Systems Regulator, APP reimbursement rules.