A fraud definition that uses the word fraudulent is not a definition.

Almost every receivables facility carries a definition of a fraudulent receivable and a covenant that the pool contains none. The covenant is usually expressed as zero tolerance and is usually not curable by equity, which is correct: a lender should not be asked to price fraud, and an originator should not be able to buy its way past it.

The definitions those covenants attach to tend to look like this:

Read as a whole, this is not a definition of fraud. It is a mechanism for one party to designate an asset as fraudulent, sitting behind two limbs that do no work because they are circular. Bolt an uncurable zero-tolerance covenant onto it and the drafting has produced something neither side actually wants: a default trigger that can be pulled by characterisation rather than by proof.

Why the usual defences do not hold

"Fraud has a settled legal meaning." It does, in the law of deceit, and that meaning requires a false representation made knowingly or recklessly with intent to induce reliance. Very little of what the market calls fraud in this asset class would survive that test, and nobody drafting these definitions intends the covenant to be limited to what could be pleaded in the High Court. The circular limbs are not importing a legal standard; they are avoiding writing one.

"Acting reasonably is a real constraint." It is a constraint on capriciousness and very little else. A lender who forms a view in good faith on incomplete information is acting reasonably even if the view turns out to be wrong. That is the right standard for a great many discretions. It is the wrong standard for a determination whose consequence is an uncurable event of default.

"The servicer's own policies keep it honest." Limb (c) makes the applicable standard a document that one party writes and can revise. If the policies are to do this work, they need to be defined, agreed at closing, and amendable only with consent. Otherwise the clause reads as though it constrains the servicer while in fact leaving the standard entirely within its gift, which serves neither party.

Why this is sharper in telecoms than elsewhere

In most trade receivables portfolios fraud is a tail event. In wholesale voice and messaging it is an operating category.

The Communications Fraud Control Association put annual losses across the sector at USD 38.95bn in its 2023 Global Fraud Loss Survey. Artificially inflated traffic, sometimes called SMS pumping, is among the fastest-growing components of that total: traffic generated to earn termination revenue rather than to reach a real recipient. Published estimates of how much international application-to-person messaging traffic is artificially inflated range from around five per cent to as much as forty per cent depending on the corridor, the sender and whose methodology is being used. That spread tells you both that the problem is large and that it is not measured consistently.

The consequence for documentation is direct. In this market, suspect traffic is not an extraordinary event that surfaces once in a facility's life. It is a routine finding, detected and dealt with continuously as part of running the book. A definition that allows any suspect traffic to be characterised as fraudulent, combined with a zero-tolerance covenant that cannot be cured, leaves the portfolio permanently one characterisation away from default. No originator can operate on that basis, and no lender genuinely wants a trigger so sensitive that it would have to decline to pull it, because a trigger nobody will use is not protection either.

The distinction that has to be drawn

The central error is collapsing three different things into one term with one consequence.

• Fraud by the seller against the lender. A fabricated receivable, an invoice for services never delivered, a forged confirmation, a related party dressed as an arm's length counterparty. This is the risk the zero-tolerance covenant exists for, and the consequence should be immediate and severe.
• Fraud in the traffic by a third party. Artificially inflated traffic injected upstream, revenue share fraud, a compromised sender. The seller may be a victim rather than a perpetrator. This is an origination quality problem, and the right remedy is that the receivable fails eligibility and is substituted out of the pool.
• Fraud by the debtor. Repudiation dressed as a dispute, or a debtor that never intended to pay. This is a credit and enforcement question and it is what the confirmation, the security and the insurance are for.

Each deserves a remedy. They are not the same remedy, and a single definition with a single uncurable consequence guarantees that at least two of the three are handled badly.

What an objective test looks like

None of the following is exotic. All of it is testable, which is the only property that matters.

• Anchor to external classifications. The industry maintains standardised fraud and dispute categories through its own forums and codes of conduct. Referencing them gives both sides a common vocabulary that neither wrote.
• Use measurable indicators. Traffic pattern deviation beyond a defined threshold against an agreed baseline; sender identifier mismatch; delivery or conversion rates outside a stated band; volume spikes on a route beyond an agreed tolerance. These are the things a fraud team actually looks at, and they can be written down.
• Fix the test date. Eligibility is tested at purchase. A separate, defined look-back window should govern what happens when something surfaces afterwards, rather than leaving the pool open to reassessment indefinitely.
• Separate the consequence from the finding. Traffic-level fraud should produce ineligibility and a defined substitution window, not a default. Seller fraud against the lender should produce an immediate event of default with no cure.
• Set a de minimis. Zero tolerance applied literally means one bad message in a pool of hundreds of millions is a breach. A materiality threshold, by value and as a proportion of the pool, is what makes zero tolerance mean something rather than everything.
• Define the policies. If the servicer's fraud and collection policies are the operative standard, they should be delivered as a condition precedent and amendable only with consent.

Keep a residual limb, but proportion its consequence

There is a real objection to all of this, and it should be met rather than ignored. Fraud is adaptive. An exhaustive list of indicators is a specification that a determined counterparty can be expected to engineer around, and a lender who has agreed to a closed list has given away the ability to respond to a method nobody had seen when the facility was signed.

The answer is not to abandon objectivity; it is to keep a residual discretion and make its consequence proportionate. A determination by the agent, acting in good faith and giving reasons, should be capable of rendering a receivable ineligible and requiring its substitution. It should not, by itself, constitute an uncurable event of default. Reserve that for the objective limbs and for seller fraud, where the finding can be evidenced.

That split gives the lender a fast remedy that works against methods nobody has seen yet, and gives the originator a facility that can survive the ordinary business of detecting fraud in a market where fraud is ordinary. It is a better bargain than either side gets from the circular version.

The uncomfortable part

An originator arguing for an objective fraud definition sounds, at first hearing, like an originator arguing for more room to have fraud in its book. That reading should be confronted rather than talked around.

The test of good faith here is what the originator asks for on the other limb. If the same party pressing for objectivity on traffic-level fraud also accepts an immediate, uncurable, no-materiality event of default for fraud by the seller against the lender, and accepts a full look-back and repurchase obligation on it, then the argument is about calibration rather than about escape. If it does not, the objection is well founded and a lender should press it.

Ask for that trade explicitly. It resolves the question quickly and in a way that neither side can misread.