diff --git a/apps/transport/lib/jobs/consolidate_bnlc_job.ex b/apps/transport/lib/jobs/consolidate_bnlc_job.ex
index 01f4788713..ad918adb9f 100644
--- a/apps/transport/lib/jobs/consolidate_bnlc_job.ex
+++ b/apps/transport/lib/jobs/consolidate_bnlc_job.ex
@@ -23,7 +23,14 @@ defmodule Transport.Jobs.ConsolidateBNLCJob do
   @s3_bucket :on_demand_validation
 
   # Custom types
-  @type consolidation_errors :: %{dataset_errors: list(), validation_errors: list(), download_errors: list()}
+  @type decode_error :: {:decode, {map(), map()}}
+  @type download_error :: {:download, {map(), map()}}
+  @type consolidation_errors :: %{
+          dataset_errors: list(),
+          validation_errors: list(),
+          download_errors: [download_error()],
+          decode_errors: [decode_error()]
+        }
   @type dataset_without_appropriate_resource_error :: %{
           error: :not_at_least_one_appropriate_resource,
           dataset_details: map()
@@ -57,11 +64,11 @@ defmodule Transport.Jobs.ConsolidateBNLCJob do
     Logger.info("Finding valid resources…")
     %{ok: resources_details, errors: validation_errors} = valid_datagouv_resources(datasets_details)
 
-    if validation_errors |> Enum.filter(&match?({:validation_error, _, _}, &1)) |> Enum.any?() do
+    if validator_unavailable?(validation_errors) do
       {:discard, "Cannot consolidate the BNLC because the TableSchema validator is not available"}
     else
       Logger.info("Downloading resources…")
-      %{ok: download_details, errors: download_errors} = download_resources(resources_details)
+      %{ok: download_details, errors: download_or_decode_errors} = download_resources(resources_details)
       Logger.info("Creating a single file")
       consolidate_resources(download_details)
 
@@ -72,13 +79,20 @@ defmodule Transport.Jobs.ConsolidateBNLCJob do
       |> send_email_recap(%{
         dataset_errors: dataset_errors,
         validation_errors: validation_errors,
-        download_errors: download_errors
+        download_errors: Enum.filter(download_or_decode_errors, &match?({:download, _}, &1)),
+        decode_errors: Enum.filter(download_or_decode_errors, &match?({:decode, _}, &1))
       })
 
       :ok
     end
   end
 
+  defp validator_unavailable?(validation_errors) do
+    validation_errors
+    |> Enum.filter(&match?({:validator_unavailable_error, _, _}, &1))
+    |> Enum.any?()
+  end
+
   defp upload_temporary_file do
     content = File.read!(@bnlc_path)
     now = DateTime.utc_now() |> DateTime.truncate(:microsecond) |> DateTime.to_string() |> String.replace(" ", "_")
@@ -128,8 +142,8 @@ defmodule Transport.Jobs.ConsolidateBNLCJob do
   end
 
   @spec format_errors(consolidation_errors()) :: binary() | nil
-  def format_errors(%{dataset_errors: _, validation_errors: _, download_errors: _} = errors) do
-    [&format_dataset_errors/1, &format_validation_errors/1, &format_download_errors/1]
+  def format_errors(%{dataset_errors: _, validation_errors: _, download_errors: _, decode_errors: _} = errors) do
+    [&format_dataset_errors/1, &format_validation_errors/1, &format_download_errors/1, &format_decode_errors/1]
     |> Enum.map_join("\n\n", fn fmt_fn -> fmt_fn.(errors) end)
     |> String.trim()
     |> case do
@@ -169,12 +183,33 @@ defmodule Transport.Jobs.ConsolidateBNLCJob do
     """
   end
 
+  @spec format_download_errors(%{download_errors: [download_error()]}) :: binary()
   def format_download_errors(%{download_errors: []}), do: ""
 
   def format_download_errors(%{download_errors: download_errors}) do
+    errors =
+      Enum.map(download_errors, fn {:download, {dataset, resource}} ->
+        {dataset, resource}
+      end)
+
     """
     <h2>Impossible de télécharger les ressources suivantes</h2>
-    #{Enum.map_join(download_errors, "\n", &link_to_resource/1)}
+    #{Enum.map_join(errors, "\n", &link_to_resource/1)}
+    """
+  end
+
+  @spec format_decode_errors(%{decode_errors: [decode_error()]}) :: binary()
+  def format_decode_errors(%{decode_errors: []}), do: ""
+
+  def format_decode_errors(%{decode_errors: decode_errors}) do
+    errors =
+      Enum.map(decode_errors, fn {:decode, {dataset, resource}} ->
+        {dataset, resource}
+      end)
+
+    """
+    <h2>Impossible de décoder les fichiers CSV suivants</h2>
+    #{Enum.map_join(errors, "\n", &link_to_resource/1)}
     """
   end
 
@@ -348,7 +383,7 @@ defmodule Transport.Jobs.ConsolidateBNLCJob do
 
   @spec valid_datagouv_resources([map()]) :: %{
           ok: [],
-          errors: [{:error, map(), map()} | {:validation_error, map(), map()}]
+          errors: [{:error, map(), map()} | {:validator_unavailable_error, map(), map()}]
         }
   @doc """
   Identifies valid resources according to the target schema.
@@ -356,14 +391,14 @@ defmodule Transport.Jobs.ConsolidateBNLCJob do
 
   Possible errors:
   - the resource is not valid according to the schema ({`:error`, _, _})
-  - the validator is not available ({`:validation_error`, _, _})
+  - the validator is not available ({`:validator_unavailable_error`, _, _})
   """
   def valid_datagouv_resources(datasets_details) do
     analyze_resource = fn dataset_details, %{"url" => resource_url} = resource ->
       case TableSchemaValidator.validate(@schema_name, resource_url) do
         %{"has_errors" => true} -> {:errors, {:error, dataset_details, resource}}
         %{"has_errors" => false} -> {:ok, {dataset_details, resource}}
-        nil -> {:errors, {:validation_error, dataset_details, resource}}
+        nil -> {:errors, {:validator_unavailable_error, dataset_details, resource}}
       end
     end
 
@@ -378,31 +413,61 @@ defmodule Transport.Jobs.ConsolidateBNLCJob do
 
   @doc """
   From a list of resource object coming from the data.gouv.fr's API, download these (valid)
-  CSV files locally and guess the CSV separator.
+  CSV files locally, guess the CSV separator and try to decode the file.
 
   The temporary download path and the guessed CSV separator are added to the resource's payload.
 
   Possible errors:
   - cannot download the resource
+  - cannot decode the CSV file
   """
-  @spec download_resources([map()]) :: %{ok: [], errors: []}
+  @spec download_resources([map()]) :: %{ok: [], errors: [decode_error() | download_error()]}
   def download_resources(resources_details) do
     resources_details
-    |> Enum.map(fn {dataset_details, %{"id" => resource_id, "url" => resource_url} = resource} ->
+    |> Enum.map(fn {dataset_details, %{"url" => resource_url} = resource} ->
       case http_client().get(resource_url, [], follow_redirect: true) do
-        {:ok, %HTTPoison.Response{status_code: 200, body: body}} ->
-          path = System.tmp_dir!() |> Path.join("consolidate_bnlc_#{resource_id}")
-          File.write!(path, body)
-          resource = Map.merge(resource, %{@download_path_key => path, @separator_key => guess_csv_separator(body)})
-          {:ok, {dataset_details, resource}}
+        {:ok, %HTTPoison.Response{status_code: 200} = response} ->
+          guess_separator_and_decode({dataset_details, resource}, response)
 
         _ ->
-          {:errors, {dataset_details, resource}}
+          {:errors, {:download, {dataset_details, resource}}}
       end
     end)
     |> normalize_ok_errors()
   end
 
+  @doc """
+  For a remote resource we successfully downloaded, we try to:
+  - guess the CSV separator of the file (using the header line)
+  - decode the CSV file with the guessed separator
+  """
+  def guess_separator_and_decode({dataset_details, %{"id" => resource_id} = resource}, %HTTPoison.Response{
+        status_code: 200,
+        body: body
+      }) do
+    path = System.tmp_dir!() |> Path.join("consolidate_bnlc_#{resource_id}")
+    File.write!(path, body)
+    resource = Map.merge(resource, %{@download_path_key => path, @separator_key => guess_csv_separator(body)})
+    check_can_decode_csv(body, {dataset_details, resource})
+  end
+
+  defp check_can_decode_csv(
+         body,
+         {dataset_details, %{@separator_key => separator, @download_path_key => path} = resource}
+       ) do
+    errors = [body] |> CSV.decode(separator: separator) |> Enum.filter(&(elem(&1, 0) == :error))
+
+    if Enum.empty?(errors) do
+      {:ok, {dataset_details, resource}}
+    else
+      # Could not decode the CSV:
+      # - we delete the temporary file since we will not include it in the consolidation
+      # - we return an error
+      File.rm!(path)
+      {:errors, {:decode, {dataset_details, resource}}}
+    end
+  end
+
   @doc """
   Make sure we always have `:ok` and `:errors` keys.
 
diff --git a/apps/transport/test/transport/jobs/consolidate_bnlc_job_test.exs b/apps/transport/test/transport/jobs/consolidate_bnlc_job_test.exs
index 10ca585d9a..a41caa9260 100644
--- a/apps/transport/test/transport/jobs/consolidate_bnlc_job_test.exs
+++ b/apps/transport/test/transport/jobs/consolidate_bnlc_job_test.exs
@@ -82,8 +82,14 @@ defmodule Transport.Test.Transport.Jobs.ConsolidateBNLCJobTest do
            %{"title" => "Ressource avec erreurs", "schema" => %{"name" => @target_schema}}}
         ],
         download_errors: [
-          {%{"page" => "https://example.com/jdd_download_error", "title" => "JDD avec erreur de téléchargement"},
-           %{"title" => "Ressource indisponible", "schema" => %{"name" => @target_schema}}}
+          {:download,
+           {%{"page" => "https://example.com/jdd_download_error", "title" => "JDD avec erreur de téléchargement"},
+            %{"title" => "Ressource indisponible", "schema" => %{"name" => @target_schema}}}}
+        ],
+        decode_errors: [
+          {:decode,
+           {%{"page" => "https://example.com/jdd_decode_error", "title" => "JDD impossible à décoder"},
+            %{"title" => "Ressource mal formatée", "schema" => %{"name" => @target_schema}}}}
         ]
       }
 
@@ -98,12 +104,22 @@ defmodule Transport.Test.Transport.Jobs.ConsolidateBNLCJobTest do
 
 
              <h2>Impossible de télécharger les ressources suivantes</h2>
-             Ressource `Ressource indisponible` (<a href="https://example.com/jdd_download_error">JDD avec erreur de téléchargement</a>)\
+             Ressource `Ressource indisponible` (<a href="https://example.com/jdd_download_error">JDD avec erreur de téléchargement</a>)
+
+
+             <h2>Impossible de décoder les fichiers CSV suivants</h2>
+             Ressource `Ressource mal formatée` (<a href="https://example.com/jdd_decode_error">JDD impossible à décoder</a>)\
              """ == ConsolidateBNLCJob.format_errors(errors)
     end
 
     test "nil when there are no errors" do
-      assert nil == ConsolidateBNLCJob.format_errors(%{dataset_errors: [], validation_errors: [], download_errors: []})
+      assert nil ==
+               ConsolidateBNLCJob.format_errors(%{
+                 dataset_errors: [],
+                 validation_errors: [],
+                 download_errors: [],
+                 decode_errors: []
+               })
     end
   end
 
@@ -154,7 +170,7 @@ defmodule Transport.Test.Transport.Jobs.ConsolidateBNLCJobTest do
     assert %{
              errors: [
                {:error, dataset_details, other_resource},
-               {:validation_error, other_dataset_details, validation_error_resource}
+               {:validator_unavailable_error, other_dataset_details, validation_error_resource}
              ],
              ok: [{dataset_details, resource}]
            } == ConsolidateBNLCJob.valid_datagouv_resources(datasets_details)
@@ -172,17 +188,28 @@ defmodule Transport.Test.Transport.Jobs.ConsolidateBNLCJobTest do
       "slug" => "foo"
     }
 
-    dataset_error_detail = %{
+    dataset_download_error_detail = %{
       "resources" => [
-        resource_error = %{
+        resource_download_error = %{
           "id" => Ecto.UUID.generate(),
           "schema" => %{"name" => @target_schema},
-          "url" => error_url = "https://example.com/other_file.csv"
+          "url" => download_error_url = "https://example.com/download_error.csv"
         }
       ],
       "slug" => "bar"
     }
 
+    dataset_decode_error_detail = %{
+      "resources" => [
+        resource_decode_error = %{
+          "id" => resource_decode_error_id = Ecto.UUID.generate(),
+          "schema" => %{"name" => @target_schema},
+          "url" => decode_error_url = "https://example.com/decode_error.csv"
+        }
+      ],
+      "slug" => "baz"
+    }
+
     Transport.HTTPoison.Mock
     |> expect(:get, fn ^url, [], [follow_redirect: true] ->
       body = """
@@ -194,14 +221,39 @@ defmodule Transport.Test.Transport.Jobs.ConsolidateBNLCJobTest do
     end)
 
     Transport.HTTPoison.Mock
-    |> expect(:get, fn ^error_url, [], [follow_redirect: true] ->
+    |> expect(:get, fn ^download_error_url, [], [follow_redirect: true] ->
       {:ok, %HTTPoison.Response{status_code: 500, body: ""}}
     end)
 
-    resources_details = [{dataset_detail, resource}, {dataset_error_detail, resource_error}]
+    Transport.HTTPoison.Mock
+    |> expect(:get, fn ^decode_error_url, [], [follow_redirect: true] ->
+      # Malformed CSV: unescaped double quotes: `""2"`
+      body = """
+      "foo","bar"
+      "1",""2"
+      """
+
+      {:ok, %HTTPoison.Response{status_code: 200, body: body}}
+    end)
+
+    resources_details = [
+      {dataset_detail, resource},
+      {dataset_download_error_detail, resource_download_error},
+      {dataset_decode_error_detail, resource_decode_error}
+    ]
 
     assert %{
-             errors: [{^dataset_error_detail, ^resource_error}],
+             errors: [
+               {:download, {^dataset_download_error_detail, ^resource_download_error}},
+               {:decode,
+                {^dataset_decode_error_detail,
+                 %{
+                   "id" => ^resource_decode_error_id,
+                   "csv_separator" => ?,,
+                   "tmp_download_path" => tmp_decode_error_download_path,
+                   "url" => ^decode_error_url
+                 }}}
+             ],
              ok: [
                {^dataset_detail,
                 %{"id" => ^resource_id, "url" => ^url, "csv_separator" => ?,, "tmp_download_path" => tmp_download_path}}
@@ -210,6 +262,7 @@ defmodule Transport.Test.Transport.Jobs.ConsolidateBNLCJobTest do
 
     assert String.ends_with?(tmp_download_path, "consolidate_bnlc_#{resource_id}")
     assert File.exists?(tmp_download_path)
+    refute File.exists?(tmp_decode_error_download_path)
   end
 
   test "guess_csv_separator" do