diff --git a/Project.toml b/Project.toml
index 74342aa99..3241a9533 100644
--- a/Project.toml
+++ b/Project.toml
@@ -95,7 +95,7 @@ RecursiveArrayTools = "3.27.2"
 Reexport = "1.0"
 ReverseDiff = "1.15.1"
 SafeTestsets = "0.1.0"
-SciMLBase = "2.103.1"
+SciMLBase = "2.117.1"
 SciMLJacobianOperators = "0.1"
 SciMLStructures = "1.3"
 SparseArrays = "1.10"
diff --git a/src/SciMLSensitivity.jl b/src/SciMLSensitivity.jl
index 7a003b4d3..619425ad4 100644
--- a/src/SciMLSensitivity.jl
+++ b/src/SciMLSensitivity.jl
@@ -83,6 +83,7 @@ include("concrete_solve.jl")
 include("second_order.jl")
 include("steadystate_adjoint.jl")
 include("sde_tools.jl")
+include("enzyme_rules.jl")
 
 export extract_local_sensitivities
 
diff --git a/src/enzyme_rules.jl b/src/enzyme_rules.jl
new file mode 100644
index 000000000..e11b6acd4
--- /dev/null
+++ b/src/enzyme_rules.jl
@@ -0,0 +1,14 @@
+# Enzyme rules for VJP choice types defined in SciMLSensitivity
+#
+# VJP choice types configure how jacobian-vector products are computed within 
+# sensitivity algorithms. They should be treated as inactive (constant) during 
+# Enzyme differentiation to prevent errors when they are stored in problem 
+# structures or other data that Enzyme differentiates through.
+#
+# Note: AbstractSensitivityAlgorithm inactive rule is handled in SciMLBase
+# to avoid type piracy.
+
+import Enzyme: EnzymeRules
+
+# VJP choice types should be inactive since they configure computation methods
+EnzymeRules.inactive_type(::Type{<:VJPChoice}) = true
diff --git a/test/enzyme_vjp_inactive.jl b/test/enzyme_vjp_inactive.jl
new file mode 100644
index 000000000..d99025896
--- /dev/null
+++ b/test/enzyme_vjp_inactive.jl
@@ -0,0 +1,72 @@
+using Test, SciMLSensitivity, Enzyme, OrdinaryDiffEq
+
+# Test that VJP choice types are treated as inactive by Enzyme  
+# The AbstractSensitivityAlgorithm inactive rule is handled in SciMLBase
+# This addresses issue #1225 where sensealg in ODEProblem constructor would fail
+
+@testset "Enzyme VJP Choice Inactive Types" begin
+
+    # Test 1: Basic test that VJP objects can be stored in data structures during Enzyme differentiation
+    @testset "VJP types in data structures" begin
+        vjp = EnzymeVJP()
+
+        function test_func(x)
+            # Store the VJP in a data structure (this would fail without inactive rules)
+            data = (value = x[1] + x[2], vjp = vjp)
+            return data.value * 2.0
+        end
+
+        x = [1.0, 2.0]
+        dx = Enzyme.make_zero(x)
+
+        # This should not throw an error
+        @test_nowarn Enzyme.autodiff(Enzyme.Reverse, test_func, Enzyme.Active, Enzyme.Duplicated(x, dx))
+        @test dx ≈ [2.0, 2.0]
+    end
+
+    # Test 2: Test different VJP choice types are inactive
+    @testset "Different VJP types inactive" begin
+        vjp_types = [EnzymeVJP(), ZygoteVJP(), ReverseDiffVJP(), TrackerVJP()]
+
+        for vjp in vjp_types
+            function test_func(x)
+                data = (value = x[1] * x[2], vjp = vjp)
+                return data.value + 1.0
+            end
+
+            x = [2.0, 3.0]
+            dx = Enzyme.make_zero(x)
+
+            @test_nowarn Enzyme.autodiff(Enzyme.Reverse, test_func, Enzyme.Active, Enzyme.Duplicated(x, dx))
+        end
+    end
+
+    # Test 3: Test sensitivity algorithms with VJP choices (integration test)
+    # Note: This test also depends on SciMLBase having AbstractSensitivityAlgorithm as inactive
+    @testset "Sensitivity algorithms with VJP choices" begin
+        function f(du, u, p, t)
+            du[1] = -p[1] * u[1]
+            du[2] = p[2] * u[2]
+        end
+
+        function loss_func(p)
+            u0 = [1.0, 2.0]
+            # Both VJP choice and sensitivity algorithm should be inactive
+            prob = ODEProblem(
+                f, u0, (0.0, 0.1), p, sensealg = BacksolveAdjoint(autojacvec = EnzymeVJP()))
+            sol = solve(prob, Tsit5())
+            return sol.u[end][1] + sol.u[end][2]
+        end
+
+        p = [0.5, 1.5]
+        dp = Enzyme.make_zero(p)
+
+        # This should not throw the "Error handling recursive stores for String" error
+        # This is the original failing case from issue #1225
+        @test_nowarn Enzyme.autodiff(Enzyme.Reverse, loss_func, Enzyme.Active, Enzyme.Duplicated(p, dp))
+
+        # Verify the gradient is computed (non-zero and finite)
+        @test all(isfinite, dp)
+        @test any(x -> abs(x) > 1e-10, dp)  # At least one component should be non-trivial
+    end
+end
diff --git a/test/runtests.jl b/test/runtests.jl
index cfb90641e..5ddde73c7 100644
--- a/test/runtests.jl
+++ b/test/runtests.jl
@@ -40,6 +40,7 @@ end
             @time @safetestset "Scalar u0" include("scalar_u.jl")
             @time @safetestset "Error Messages" include("error_messages.jl")
             @time @safetestset "Autodiff Events" include("autodiff_events.jl")
+            @time @safetestset "Enzyme VJP Inactive" include("enzyme_vjp_inactive.jl")
         end
     end