diff --git a/NEWS.md b/NEWS.md
index 78f412ac8..5627e18ed 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -1,5 +1,7 @@
 # parsnip (development version)
 
+* We no longer add `eval_time` arguments to the prediction specification for the engine (#1039).
+
 * parsnip now lets the engines for [mlp()] check for acceptable values of the activation function (#1019)
 
 * Tightened logic for outcome checking. This resolves issues—some errors and some silent failures—when atomic outcome variables have an attribute (#1060, #1061).
diff --git a/R/predict_hazard.R b/R/predict_hazard.R
index 84b49e13a..fe75b5007 100644
--- a/R/predict_hazard.R
+++ b/R/predict_hazard.R
@@ -33,7 +33,6 @@ predict_hazard.model_fit <- function(object,
     new_data <- object$spec$method$pred$hazard$pre(new_data, object)
 
   # Pass some extra arguments to be used in post-processor
-  object$spec$method$pred$hazard$args$eval_time <- eval_time
   pred_call <- make_pred_call(object$spec$method$pred$hazard)
 
   res <- eval_tidy(pred_call)
diff --git a/R/predict_survival.R b/R/predict_survival.R
index fbfe4eefe..cbe4a86fe 100644
--- a/R/predict_survival.R
+++ b/R/predict_survival.R
@@ -35,7 +35,6 @@ predict_survival.model_fit <- function(object,
     new_data <- object$spec$method$pred$survival$pre(new_data, object)
 
   # Pass some extra arguments to be used in post-processor
-  object$spec$method$pred$survival$args$eval_time <- eval_time
   pred_call <- make_pred_call(object$spec$method$pred$survival)
 
   res <- eval_tidy(pred_call)