Sirion Acquires AI Pioneer Eigen Technologies to Accelerate Document AI Capabilities.

New analysis shows ARRC-recommended LIBOR fallback provisions in loan and credit agreements still create conflict with other provisions

In this article Friedman Kaplan attorneys Asaf Reindel, Stan Chiueh, and Jessica Julich share findings from a review of the LIBOR fallback and other provisions in 300 recent loan and credit agreements, which were extracted using our natural language processing platform.

This article was originally published on Friedman Kaplan's website on 19 August 2020.

Friedman Kaplan and Eigen Technologies analysis shows ARRC-recommended LIBOR fallback provisions in loan and credit agreements still create conflict with other provisions

In a previous article, published in September 2019, we identified a standard loan or credit agreement provision that seems to conflict with the LIBOR fallback provisions recommended by the Alternative Reference Rates Committee (ARRC). That “cost of funding” provision typically permits a lender to (among other things) suspend its obligation to make or continue to fund LIBOR-based loans under the agreement if the lender determines that LIBOR does not adequately and fairly reflect its cost of funding.

When such a provision is left unchanged in a loan or credit agreement into which the ARRC recommended fallback provision has been incorporated, this potentially gives the lender a free option to abandon LIBOR at any time (and possibly switch to another rate such as the Prime Rate), without triggering the LIBOR fallback provision. That option would exist because even today the market of unsecured interbank lending underlying LIBOR is very thin, and is getting thinner. In other words, LIBOR arguably does not reflect a lender’s cost of funding right now.

In this article, we present the result of a review we have conducted with Eigen Technologies, a natural language processing (NLP) provider whose technology is used across the banking, financial services, legal, insurance, and management consulting sectors to unlock qualitative data found in contracts and other written documents. Eigen’s no-code AI platform was used to extract data from 300 loan and credit agreements that have been filed publicly with the SEC since September 2019. Using Eigen to conduct multiple scanning steps for provisions we identified (each step informed by previous scans), we analyzed whether the internal potential conflict we have identified persists, and whether market participants have attempted to address it.

As further described below, we found that most of these agreements (67%) do not address the conflict we identified previously, a small minority (16.7%) avoid the conflict successfully, and the others (16.3%) address it in a manner that in our view is likely to be insufficient or problematic.


In our September 2019 article, we noted that in recent years LIBOR has become increasingly disconnected from its underlying market. Since then, unsecured interbank lending activity has continued to dwindle. This was reiterated recently by Andrew Bailey, Governor of the Bank of England, in a speech given on July 13, 2020:

So the paradox of Libor is that if we are looking for a robust way to create transparency on bank funding costs, Libor is not that rate. Because the market it measures is a small to disappearing part of overall bank funding, it can no longer claim to accurately reflect the marginal cost of funds for banks, nor to provide end users with confidence their interest payments are directly linked to those costs.

Governor Bailey went on to observe that “overall, this leaves Libor rates capturing (at best) only a tiny fraction of overall funding costs even in benign market conditions.”

It therefore is becoming increasingly important to avoid the type of conflict we previously identified between LIBOR fallback provisions and cost of funding provisions in loan and credit agreements.


With the power of Eigen’s NLP machine learning platform, we determined that out of 300 publicly filed agreements examined, nearly all of them (298, or 99.3%) include the ARRC’s recommended LIBOR fallback provisions, while the remaining two use other LIBOR fallback language. Out of those 300 agreements, we determined that 252 include a potentially conflicting cost of funding provision. Of these:

  • 49 agreements provide that the cost of funding provision was “subject to” the LIBOR fallback provision [1] (or, in a small number of agreements, by providing similarly that the cost of funding provision may only be used “unless and until a Benchmark Replacement is implemented in accordance with [the fallback provisions]”). As further discussed below, in our view this does not fully address the conflict.
  • Two agreements address the potential conflict in a unique (and more successful) way, as described below.
  • 48 agreements do not include a cost of funding provision, which likely was done intentionally to avoid this potential conflict.

That leaves 201 agreements – two thirds of the total – with potentially conflicting provisions.


As noted, about half of the agreements where the parties seem to have been aware of the potential conflict between a LIBOR fallback provision and a cost of funding provision address it by making the cost of funding provision expressly “subject to” the LIBOR fallback provision. In our view, this approach resolves the conflict only partially.

This approach makes it clear that once the LIBOR fallback provision has been triggered, the lender may no longer invoke the suspension right under the cost of funding provision. This arguably prevents a lender from having two possible routes to LIBOR replacement after the fallback provision is triggered.

However, this approach still does not prevent a lender from suspending LIBOR-based lending pursuant to the cost of funding provision before the LIBOR fallback provision is triggered, which a lender arguably could do even today, given LIBOR’s precarious situation relative to the bank funding markets. In other words, a lender could suspend LIBOR-based lending without regard to the agreed-upon LIBOR fallback process, at least before that process is triggered.

That may be advantageous and even intentional from a lender’s perspective, but it is not clear that borrower necessarily understand the implications. Indeed, it seems unlikely that an informed borrower would intentionally give its lender such an early trigger option when the parties also have agreed on a LIBOR fallback provision (which typically already includes an early trigger option).

The other way in which this conflict seems to be addressed is by eliminating the cost of funding provision when the fallback provision is added. We saw this approach used in 16.7% of the agreements in our review.

This is the cleanest and most certain way to eliminate the conflict, since the lender would not be able to suspend LIBOR-based lending at any time based on the argument that LIBOR does not reflect its cost of funding.

Another approach taken in two agreements we examined reaches a largely similar result in a different way, without removing the cost of funding provision. Under these agreements, a lender’s determination that LIBOR does not adequately and fairly reflect its cost of funding itself triggers the LIBOR fallback provision. These agreements also provide that a regulatory determination that LIBOR is “no longer representative” (which typically would trigger the LIBOR fallback provision) also would constitute a circumstance in which LIBOR no longer reflects the lender’s cost of funding. [2]

For a lender that is reluctant to remove the cost of funding provision from the agreement, this may be a useful approach that avoids the “parallel track” problem described above. One difference between this approach and the elimination of the cost of funding provision from the agreement, is that this approach gives the lender a right to trigger the LIBOR fallback provision earlier than it could under the ARRC-recommended fallback provision. Since triggering the fallback process very early could create uncertainty as to the replacement benchmark, in our view removing the cost of funding provision is a better solution to resolve this conflict.


Parties that wish to avoid a conflict between a cost of funding provision and a LIBOR fallback provision would be well advised to remove the cost of funding provision from the agreement entirely. This would avoid both of potential problems we have identified: a lender would not be able to suspend LIBOR-based lending before the LIBOR fallback provision is triggered based on a determination that LIBOR does not adequately and fairly reflect its cost of funding, and it would also not be able to do so after the LIBOR fallback provision is triggered.

This article is not intended to provide legal advice for any specific situation, and no legal or business decision should be based on its content. If you would like advice on your specific situation, please feel free to contact Friedman Kaplan.


[1] These agreements typically achieve this by prefacing the cost of funding and related provisions with the phrase “subject to [the fallback provisions]”.

[2] An example of such a contractual provision is as follows:

“(b) Notwithstanding the foregoing or anything to the contrary in this Agreement or any other Loan Document, if the Administrative Agent determines (which determination shall be conclusive absent manifest error), or the Required Lenders notify the Administrative Agent (with a copy to the Borrower) that the Required Lenders have determined, that any one or more of the following (each, a “ Benchmark Transition Event ”) has occurred:

(i) the circumstances set forth in [the cost of funding provision] have arisen (including, without limitation, a public statement or publication of information by the regulatory supervisor for the administrator of the LIBO Rate described in clause (ii) of this Section (b) announcing that the LIBO Rate is no longer representative) and such circumstances are unlikely to be temporary,

then the Administrative Agent and the Borrower may amend this Agreement to replace the LIBO Rate with a Benchmark Replacement…”