The most dangerous words in a bank guarantee are not the unclear ones. They are the words that quietly change what the instrument is.
A guarantee can look entirely intact — the right title, a bank's signature, a payment undertaking — and still have been turned into something else by a single inserted clause. "Upon proof of default." "Subject to the contract." "Provided the breach is established to the bank's satisfaction." Each phrase sounds reasonable. Each was probably added to make the instrument fairer or safer. And each can pull the guarantee away from the one thing it exists to do: pay against a complying demand.
A word on terms. Independence does not mean a demand explains nothing. Under the URDG 758, a demand must ordinarily be accompanied by a supporting statement indicating in what respect the applicant is in breach, unless the guarantee expressly dispenses with it (URDG 758, article 15(a)). Independence is not silence, and it is not the absence of documents. It is a narrower idea: the guarantee stands apart from the underlying contract, and the bank pays against what the guarantee requires to be presented — not against its own findings about the contract's performance.
Where a reasonable-sounding condition leads depends on which of two very different things it does — and the two consequences are worth keeping apart.
First, the condition that the rules can simply set aside. If the undertaking remains a demand guarantee governed by URDG 758, a condition that specifies no document, no index, and nothing determinable from the guarantor's own records by which its fulfilment is to be judged is treated as non-documentary and is disregarded (URDG 758, article 7). A loosely drafted performance condition of this kind does not turn the guarantor into a judge of the contract; more often, it simply falls away. The examination task is unchanged.
Second, the condition that changes the character of the instrument. Wording can go further — tying the bank's liability to the actual establishment of the applicant's default, performance, or loss. Here the problem is not article 7 disregard. An undertaking whose payment truly depends on proving underlying default may no longer be an independent demand guarantee at all: it edges toward an accessory or conditional undertaking — a different legal animal — which URDG 758 may not govern as a demand guarantee. The danger is not that the bank must examine more; it is that the parties no longer hold the instrument they thought they had.
So there are three outcomes a clause can produce, not one: a documentary requirement the bank examines on its face; a non-documentary condition the rules disregard; and, at the extreme, wording that changes what the undertaking legally is. Only the first two live comfortably inside a demand guarantee. The third is the quietest and the most consequential — and the one drafters least often realise they have written.
The shift can be invisible in drafting and dramatic in enforcement. The beneficiary who expected liquidity discovers the instrument now demands proof. The applicant who expected protection discovers the bank may still pay on a bare statement. The bank that expected documentary certainty finds the wording has imported facts, arguments, and the whole contractual history into what was meant to be a documentary undertaking.
The point is not that conditional wording is always wrong, nor that every added document weakens a guarantee. A well-chosen documentary condition can make an instrument more precise. The danger lies in two particular places — the condition the rules will ignore, leaving a party relying on protection that was never there; and the condition that quietly converts an independent guarantee into something else.
This is why drafting a guarantee is not an exercise in legal elegance. It is an exercise in operational truth. Before issuance, two questions matter, not one: what task does this wording require the bank to perform when a demand is presented — and does this wording still describe a demand guarantee at all?
So the most dangerous words are not the harshest, the broadest, or the most technical. They are the quiet ones that change the bank's task — and, in the extreme, change the very nature of the undertaking.
A guarantee draws its strength from the narrowness of the question it puts to the bank, and from being the instrument the parties believe it to be. The most dangerous words quietly unsettle one or the other.