Your loss when a counterparty defaults is the replacement cost of the trade (the positive mark-to-market), not the notional. Collateral and netting shrink it further.
On a derivative the other side can default any time the trade has positive value to you. You then have to replace the hedge at market, so your exposure is max(V, 0), a small fraction of the notional. Move the mark-to-market, the future-exposure add-on, and the collateral, and watch the exposure bar against the notional.
EAD = max( CE + add-on − collateral, 0 )
EAD = max( A$8.0m + A$2.0m − A$0.00m, 0 ) = A$10.0m
EL = PD · LGD · EAD = 2.00% · 60.0% · A$10.0m = A$0.12m
With a Credit Support Annex, posted collateral nets against the exposure. Without one, collateral does not apply and you are an unsecured creditor.
A US$100 million notional swap can produce a US$2 to US$5 million loss rather than a US$100 million loss. Why is the notional the wrong number for credit risk, and how do daily variation margin under a CSA and central clearing each push the exposure at default toward zero? When does the future-exposure add-on still matter even after today's mark-to-market is fully collateralised?
Assumption: a single-trade, single-period view in the Hull (2022) §24 spirit. EAD = max(CE + PFE add-on − collateral, 0), with the add-on a stylised percent of notional rather than a full simulated potential future exposure. PD and LGD are taken over one horizon.