---
title: give without strings and still see where the money goes
canonical_url: https://ensurance.app/guide/give-without-strings-and-see-where-the-money-goes
markdown_url: https://ensurance.app/guide/give-without-strings-and-see-where-the-money-goes.md
subtitle: "cash transfer fraud is fought at enrollment, at payment and afterward. when the recipient has nothing to qualify for, the check comes afterward, on a public record. here is what to read"
category: nature-finance
---

# give without strings and still see where the money goes

*cash transfer fraud is fought at enrollment, at payment and afterward. when the recipient has nothing to qualify for, the check comes afterward, on a public record. here is what to read*

Cash transfer fraud is money meant for a recipient that is lost to a false claim, a diverted payment or a bribe. Programs fight it in three places: who gets enrolled, how the payment is delivered, and whether anyone checks after it arrives. The best of them publish their losses, and none gets to zero.

This post is for the funder who wants to give without strings and still see where the money went. It covers how cash programs handle fraud, what anyone can read on an account in [**ensurance**](/?from=guide), and where that record stops.

## how programs handle cash transfer fraud today

A 2014 World Bank course on social protection splits the problem three ways. **Error** is unintentional, by the claimant or by an official. **Fraud** is a claimant's intentional deceit. **Corruption** is staff abusing their position. Citing a 2006 benchmark by the UK National Audit Office, it puts fraud and error at 2 to 5% of social protection spending in high-income countries, and 5 to 10% in means-tested, income-replacement and disability programs. Those figures are old. Its advice: prevent, detect and deter, but do not "scare good-standing beneficiaries."

In that benchmark, the programs with the most rules about who qualifies show the higher rates, error included. Every eligibility rule is something to get wrong or to lie about. Fewer checks are not safer, though: GiveDirectly reports that a one-touch enrollment model in Malawi saw fraudulent enrollment reach 9.8% before added verification brought it to 0.8%.

GiveDirectly, which sends cash with no conditions on how it is spent, publishes its own numbers. Its 2025 risk report says known fraud losses totaled 0.27% in 2025, up from 0.19% in 2024, against a target of under 0.5%. Its audit team followed up directly with 17.3% of recipients to confirm safe receipt and investigated more than 22,000 fraud flags, most of which did not turn out to be losses. One 2025 program in the Democratic Republic of Congo, run with remote targeting and payment, produced a 2.98% leakage rate, and GiveDirectly says it has redesigned the next round.

GiveDirectly has published a larger, earlier loss in the same country. In 2022, staff in its Congo program registered payment SIM cards in recipients' names, kept them, and handed recipients different ones. Internal audit found it in January 2023, five months after the first payments were diverted. GiveDirectly's update puts the loss at just under $1.2 million, 0.8% of the $144 million it delivered globally that year. It paused payments, restarted them in February 2024, began paying back the 1,900 families whose first five months were diverted, changed who registers SIMs and added an outside whistleblowing line.

The loss happened in delivery, not in what recipients did with the money. It was caught by auditors, not by an eligibility rule: new internal audit staff saw signs of fraud in a neighboring program and traced it back. And the program could pay families back because it runs the payments.

## where the condition lives in fraud control

Cash programs are right about the hard parts: conditions cost money to police, people who live in a place know what it needs better than a form does, and the best programs answer fraud with an audit trail, not with more rules for the recipient.

The condition does not disappear. It moves to enrollment: a list of who qualifies, policed by an office. That list is where the means-tested error rates come from, and in the Congo case the theft sat one step later, with the staff who registered payment SIMs.

Ensurance goes one step further: no condition on the recipient, and a public record for whoever pays. The record lets the next payment go elsewhere; it does not stop one already made.

## what a funder can read on any agent account

In ensurance, a place, a purpose or a person can be represented by an **agent**, and every agent has an account. Money reaches those accounts in two ways. Someone buys a certificate that names the agent, which is **direct funding**. Or the protocol's **splits**, public contracts that divide incoming money by fixed shares, route **ensurance proceeds** to it. We call proceeds **universal BASIN income**, or uBi, because the account does not have to hold, do or qualify for anything to receive them. No amount is promised to any account.

The watershed, the forest and the rangeland exist whether or not anyone buys anything, and so do the species on them and the people who keep them. Ensurance is how money reaches them. The account is not the place, and what it receives is not what the place is worth. The order we design for puts the land first: a watershed's own account before its species and flows, and those before the people who keep them. The live record shows what the splits route today.

With no check before the money lands, the check comes after, and it has to be public. Here is what anyone can read today, with no login, against any agent account:

| what you read | where it comes from | what it does not prove |
|---|---|---|
| **Claims**: the account's stated purpose and place | Set by whoever keeps the account | Anything, until evidence matches it |
| **Flows**: the splits that route proceeds and who they name | Onchain split contracts, shown at [/proceeds](/proceeds?from=guide) | What the money was then used for |
| **Holdings**: what the account owns | The account's onchain balances | Why it holds them |
| **Activity**: recent transfers and program runs | Onchain history and the app's program log | That the work on the ground was done |
| **Place**: a linked watershed, ecoregion or parcel | The app's GIS records | Who holds the ground |
| **Bound readings**: a dated reading with a unit and a method, tied to a place the account claims | The observation record | A trend, or what happens next |

The agent page lists these checks as met, partial or a gap, each with what was observed. Flows are read separately, at /proceeds. There is no trust score. Two sources that disagree both stay on the record.

Here is what that record lets you see on one account. When we read the [Roaring Fork River watershed account](/roaring-fork-river.basin?from=guide) in Colorado on 17 November 2026, its purpose and place were set, it held tokens and showed recent transfers and program runs, and it watched one indicator, water flow. Its page also listed five gaps: no GIS geography linked, one ensurance holding that says nothing about its stated purpose, no onchain identity registration, no attestations from anyone else, and no reading of the flow it watches. A gap is not an accusation. It is a funder's question, already on the page. Open it now; it may have changed.

[What MRV actually is](/guide/what-mrv-actually-is?from=guide) shows how one reading serves as measurement, reporting and verification; [the reading on the account is the rule](/guide/the-reading-on-the-account-is-the-rule?from=guide) shows how a reading could move money set aside for a place.

## the limits of the record

- **Nothing claws back proceeds already paid.**
- **Nothing freezes an account.** No reading, gap or attestation stops an account from receiving what its splits route to it.
- **No office makes an account whole.** GiveDirectly could pay families back because it runs the payments. Splits have no one in the middle to divert a payment, and no one whose job is to repay. Whoever controls an account controls what lands in it.
- **Empty does not mean clean.** A read with no rows means nothing was found in what was read. It does not mean nothing happened.

What the record does is let those who pay see a problem and stop. A funder can stop buying a certificate that names the account; that is direct funding, theirs to withhold. Proceeds keep following the splits, which no single buyer switches off. The next dollar can go to an account whose record holds up.

:::johnson
**the check comes after the payment, and anyone can run it.**

Nothing is asked of the account. Those who pay read the record and decide whether to keep paying.

[read the series' last post: open an account for the place you keep →](/guide/open-an-account-for-the-place-you-keep?from=guide)
:::

## how to check an account, in four steps

1. **Read the claim.** Open the agent page. Note the purpose and the place it names.
2. **Read the flows.** On [/proceeds](/proceeds?from=guide), follow the splits from the origin, where proceeds start. See whether one names the account, and what that split has distributed.
3. **Read what it holds, does and measures.** Do transfers show money moving or sitting? Is a watershed or parcel linked? Is there a dated, bound reading of what the account says it watches?
4. **Decide your next payment.** Keep funding, ask the keeper about the gap, or fund another account. That decision is the only condition.

## frequently asked questions

### how do cash transfer programs prevent fraud?

They control enrollment, secure delivery and follow up after payment; a World Bank course frames it as prevent, detect and deter without scaring honest recipients. No program reports zero loss.

### can you verify where conservation money goes?

For proceeds, yes, up to the account: the splits are onchain, and an agent's holdings, transfers and place link are public. What happens on the ground needs a bound reading or an independent source, and the agent page shows whether an account has either.

### what happens if an account lies?

The gap stays on the page for anyone to see. Nothing claws back what was paid or freezes the account. Those who pay can stop paying.

## next step

Read the last post in the series: [open an account for the place you keep](/guide/open-an-account-for-the-place-you-keep?from=guide). If you fund monitoring across a portfolio, [foundations](/solutions/foundations?from=guide&topic=universal-basin-income) shows how a funder pays for readings that report to a board.

## the series

Universal BASIN income: an account for every place, purpose and person. Six posts.

1. [what a universal basic income for nature would take](/guide/what-a-universal-basic-income-for-nature-would-take?from=guide): start here
2. [unconditional cash transfers always have a condition](/guide/unconditional-cash-transfers-always-have-a-condition?from=guide): where the condition hides
3. [what the permanent fund dividend teaches conservation](/guide/what-the-permanent-fund-dividend-teaches-conservation?from=guide): pay from the earnings, keep the principal
4. [two payments people call a conservation dividend](/guide/direct-funding-proceeds-and-distributions-are-three-payments?from=guide): which money goes to whom
5. [give without strings and still see where the money goes](/guide/give-without-strings-and-see-where-the-money-goes?from=guide): the record anyone can read
6. [open an account for the place you keep](/guide/open-an-account-for-the-place-you-keep?from=guide): how to open one

## sources

[GiveDirectly, risk report 2025](https://www.givedirectly.org/risk-report-2025) — known fraud losses 0.27% (0.19% the year before; target under 0.5%); 17.3% of recipients followed up by audit; over 22,000 fraud flags investigated; 2.98% leakage in one DRC program; fraudulent enrollment of 9.8%, then 0.8%, in Malawi

[GiveDirectly, DRC case 2023 and updates](https://www.givedirectly.org/drc-case-2023/) — the SIM registration scheme from late August 2022, found by internal audit in January 2023; just under $1.2M diverted, 0.8% of $144M delivered globally; payments paused and restarted in February 2024; 1,900 families being paid back (update of September 2024); controls changed

[Emil Daniel Tesliuc, World Bank, *Reducing Fraud, Error and Corruption (EFC) in Social Protection Programs*, Social Safety Net Core Course, December 2014](https://www.worldbank.org/content/dam/Worldbank/Event/social-protection/Tesliuc%20-%20Fraud%20and%20corruption.pdf) — error, fraud and corruption by intent and actor; 2–5% of social protection spending in high-income countries, 5–10% in means-tested, income-replacement and disability programs (citing the UK National Audit Office, 2006); prevent, detect, deter; "do not scare good-standing beneficiaries"
