diff --git a/spec/03-plugins/38-ai-proxy/03-anthropic_integration_spec.lua b/spec/03-plugins/38-ai-proxy/03-anthropic_integration_spec.lua
index dd52f7e066a9..6d87425054e8 100644
--- a/spec/03-plugins/38-ai-proxy/03-anthropic_integration_spec.lua
+++ b/spec/03-plugins/38-ai-proxy/03-anthropic_integration_spec.lua
@@ -218,6 +218,35 @@ for _, strategy in helpers.all_strategies() do if strategy ~= "cassandra" then
           },
         },
       }
+
+      local chat_good_no_allow_override = assert(bp.routes:insert {
+        service = empty_service,
+        protocols = { "http" },
+        strip_path = true,
+        paths = { "/anthropic/llm/v1/chat/good-no-allow-override" }
+      })
+      bp.plugins:insert {
+        name = PLUGIN_NAME,
+        route = { id = chat_good_no_allow_override.id },
+        config = {
+          route_type = "llm/v1/chat",
+          auth = {
+            header_name = "x-api-key",
+            header_value = "anthropic-key",
+            allow_auth_override = false,
+          },
+          model = {
+            name = "claude-2.1",
+            provider = "anthropic",
+            options = {
+              max_tokens = 256,
+              temperature = 1.0,
+              upstream_url = "http://"..helpers.mock_upstream_host..":"..MOCK_PORT.."/llm/v1/chat/good",
+              anthropic_version = "2023-06-01",
+            },
+          },
+        },
+      }
       --
 
       -- 200 chat bad upstream response with one option
@@ -551,6 +580,105 @@ for _, strategy in helpers.all_strategies() do if strategy ~= "cassandra" then
         }, json.choices[1].message)
       end)
 
+      it("good request with client right header auth", function()
+        local r = client:get("/anthropic/llm/v1/chat/good", {
+          headers = {
+            ["content-type"] = "application/json",
+            ["accept"] = "application/json",
+            ["x-api-key"] = "anthropic-key",
+          },
+          body = pl_file.read("spec/fixtures/ai-proxy/anthropic/llm-v1-chat/requests/good.json"),
+        })
+
+        local body = assert.res_status(200 , r)
+        local json = cjson.decode(body)
+
+        -- check this is in the 'kong' response format
+        -- assert.equals(json.id, "chatcmpl-8T6YwgvjQVVnGbJ2w8hpOA17SeNy2")
+        assert.equals(json.model, "claude-2.1")
+        assert.equals(json.object, "chat.content")
+        assert.equals(r.headers["X-Kong-LLM-Model"], "anthropic/claude-2.1")
+
+        assert.is_table(json.choices)
+        assert.is_table(json.choices[1].message)
+        assert.same({
+          content = "The sum of 1 + 1 is 2.",
+          role = "assistant",
+        }, json.choices[1].message)
+      end)
+
+      it("good request with client wrong header auth", function()
+        local r = client:get("/anthropic/llm/v1/chat/good", {
+          headers = {
+            ["content-type"] = "application/json",
+            ["accept"] = "application/json",
+            ["x-api-key"] = "wrong",
+          },
+          body = pl_file.read("spec/fixtures/ai-proxy/anthropic/llm-v1-chat/requests/good.json"),
+        })
+
+        local body = assert.res_status(401 , r)
+        local json = cjson.decode(body)
+
+        -- check this is in the 'kong' response format
+        assert.is_truthy(json.error)
+        assert.equals(json.error.type, "authentication_error")
+      end)
+
+      it("good request with client right header auth and no allow_auth_override", function()
+        local r = client:get("/anthropic/llm/v1/chat/good-no-allow-override", {
+          headers = {
+            ["content-type"] = "application/json",
+            ["accept"] = "application/json",
+            ["x-api-key"] = "anthropic-key",
+          },
+          body = pl_file.read("spec/fixtures/ai-proxy/anthropic/llm-v1-chat/requests/good.json"),
+        })
+
+        local body = assert.res_status(200 , r)
+        local json = cjson.decode(body)
+
+        -- check this is in the 'kong' response format
+        -- assert.equals(json.id, "chatcmpl-8T6YwgvjQVVnGbJ2w8hpOA17SeNy2")
+        assert.equals(json.model, "claude-2.1")
+        assert.equals(json.object, "chat.content")
+        assert.equals(r.headers["X-Kong-LLM-Model"], "anthropic/claude-2.1")
+
+        assert.is_table(json.choices)
+        assert.is_table(json.choices[1].message)
+        assert.same({
+          content = "The sum of 1 + 1 is 2.",
+          role = "assistant",
+        }, json.choices[1].message)
+      end)
+
+      it("good request with client wrong header auth and no allow_auth_override", function()
+        local r = client:get("/anthropic/llm/v1/chat/good-no-allow-override", {
+          headers = {
+            ["content-type"] = "application/json",
+            ["accept"] = "application/json",
+            ["x-api-key"] = "wrong",
+          },
+          body = pl_file.read("spec/fixtures/ai-proxy/anthropic/llm-v1-chat/requests/good.json"),
+        })
+
+        local body = assert.res_status(200 , r)
+        local json = cjson.decode(body)
+
+        -- check this is in the 'kong' response format
+        -- assert.equals(json.id, "chatcmpl-8T6YwgvjQVVnGbJ2w8hpOA17SeNy2")
+        assert.equals(json.model, "claude-2.1")
+        assert.equals(json.object, "chat.content")
+        assert.equals(r.headers["X-Kong-LLM-Model"], "anthropic/claude-2.1")
+
+        assert.is_table(json.choices)
+        assert.is_table(json.choices[1].message)
+        assert.same({
+          content = "The sum of 1 + 1 is 2.",
+          role = "assistant",
+        }, json.choices[1].message)
+      end)
+
       it("bad upstream response", function()
         local r = client:get("/anthropic/llm/v1/chat/bad_upstream_response", {
           headers = {
diff --git a/spec/03-plugins/38-ai-proxy/04-cohere_integration_spec.lua b/spec/03-plugins/38-ai-proxy/04-cohere_integration_spec.lua
index eb52249c8fb4..d3d0f55a9ce9 100644
--- a/spec/03-plugins/38-ai-proxy/04-cohere_integration_spec.lua
+++ b/spec/03-plugins/38-ai-proxy/04-cohere_integration_spec.lua
@@ -166,6 +166,33 @@ for _, strategy in helpers.all_strategies() do if strategy ~= "cassandra" then
           },
         },
       }
+      local chat_good_no_allow_override = assert(bp.routes:insert {
+        service = empty_service,
+        protocols = { "http" },
+        strip_path = true,
+        paths = { "/cohere/llm/v1/chat/good-no-allow-override" }
+      })
+      bp.plugins:insert {
+        name = PLUGIN_NAME,
+        route = { id = chat_good_no_allow_override.id },
+        config = {
+          route_type = "llm/v1/chat",
+          auth = {
+            header_name = "Authorization",
+            header_value = "Bearer cohere-key",
+            allow_auth_override = false,
+          },
+          model = {
+            name = "command",
+            provider = "cohere",
+            options = {
+              max_tokens = 256,
+              temperature = 1.0,
+              upstream_url = "http://"..helpers.mock_upstream_host..":"..MOCK_PORT.."/llm/v1/chat/good",
+            },
+          },
+        },
+      }
       --
 
       -- 200 chat bad upstream response with one option
@@ -426,6 +453,102 @@ for _, strategy in helpers.all_strategies() do if strategy ~= "cassandra" then
         }, json.choices[1].message)
       end)
 
+      it("good request with right client auth", function()
+        local r = client:get("/cohere/llm/v1/chat/good", {
+          headers = {
+            ["content-type"] = "application/json",
+            ["accept"] = "application/json",
+            ["Authorization"] = "Bearer cohere-key",
+          },
+          body = pl_file.read("spec/fixtures/ai-proxy/cohere/llm-v1-chat/requests/good.json"),
+        })
+
+        local body = assert.res_status(200 , r)
+        local json = cjson.decode(body)
+
+        -- check this is in the 'kong' response format
+        assert.equals(json.model, "command")
+        assert.equals(json.object, "chat.completion")
+        assert.equals(r.headers["X-Kong-LLM-Model"], "cohere/command")
+
+        assert.is_table(json.choices)
+        assert.is_table(json.choices[1].message)
+        assert.same({
+          content = "The sum of 1 + 1 is 2.",
+          role = "assistant",
+        }, json.choices[1].message)
+      end)
+
+      it("good request with wrong client auth", function()
+        local r = client:get("/cohere/llm/v1/chat/good", {
+          headers = {
+            ["content-type"] = "application/json",
+            ["accept"] = "application/json",
+            ["Authorization"] = "Bearer wrong",
+          },
+          body = pl_file.read("spec/fixtures/ai-proxy/cohere/llm-v1-chat/requests/good.json"),
+        })
+
+        local body = assert.res_status(401 , r)
+        local json = cjson.decode(body)
+
+        -- check this is in the 'kong' response format
+        assert.is_truthy(json.message)
+        assert.equals(json.message, "invalid api token")
+      end)
+
+      it("good request with right client auth and no allow_auth_override", function()
+        local r = client:get("/cohere/llm/v1/chat/good-no-allow-override", {
+          headers = {
+            ["content-type"] = "application/json",
+            ["accept"] = "application/json",
+            ["Authorization"] = "Bearer cohere-key",
+          },
+          body = pl_file.read("spec/fixtures/ai-proxy/cohere/llm-v1-chat/requests/good.json"),
+        })
+
+        local body = assert.res_status(200 , r)
+        local json = cjson.decode(body)
+
+        -- check this is in the 'kong' response format
+        assert.equals(json.model, "command")
+        assert.equals(json.object, "chat.completion")
+        assert.equals(r.headers["X-Kong-LLM-Model"], "cohere/command")
+
+        assert.is_table(json.choices)
+        assert.is_table(json.choices[1].message)
+        assert.same({
+          content = "The sum of 1 + 1 is 2.",
+          role = "assistant",
+        }, json.choices[1].message)
+      end)
+
+      it("good request with wrong client auth and no allow_auth_override", function()
+        local r = client:get("/cohere/llm/v1/chat/good-no-allow-override", {
+          headers = {
+            ["content-type"] = "application/json",
+            ["accept"] = "application/json",
+            ["Authorization"] = "Bearer wrong",
+          },
+          body = pl_file.read("spec/fixtures/ai-proxy/cohere/llm-v1-chat/requests/good.json"),
+        })
+
+        local body = assert.res_status(200 , r)
+        local json = cjson.decode(body)
+
+        -- check this is in the 'kong' response format
+        assert.equals(json.model, "command")
+        assert.equals(json.object, "chat.completion")
+        assert.equals(r.headers["X-Kong-LLM-Model"], "cohere/command")
+
+        assert.is_table(json.choices)
+        assert.is_table(json.choices[1].message)
+        assert.same({
+          content = "The sum of 1 + 1 is 2.",
+          role = "assistant",
+        }, json.choices[1].message)
+      end)
+
       it("bad upstream response", function()
         local r = client:get("/cohere/llm/v1/chat/bad_upstream_response", {
           headers = {
diff --git a/spec/03-plugins/38-ai-proxy/05-azure_integration_spec.lua b/spec/03-plugins/38-ai-proxy/05-azure_integration_spec.lua
index 82385720efcf..baa6a618389d 100644
--- a/spec/03-plugins/38-ai-proxy/05-azure_integration_spec.lua
+++ b/spec/03-plugins/38-ai-proxy/05-azure_integration_spec.lua
@@ -168,6 +168,36 @@ for _, strategy in helpers.all_strategies() do if strategy ~= "cassandra" then
           },
         },
       }
+
+      local chat_good_no_allow_override = assert(bp.routes:insert {
+        service = empty_service,
+        protocols = { "http" },
+        strip_path = true,
+        paths = { "/azure/llm/v1/chat/good-no-allow-override" }
+      })
+      bp.plugins:insert {
+        name = PLUGIN_NAME,
+        route = { id = chat_good_no_allow_override.id },
+        config = {
+          route_type = "llm/v1/chat",
+          auth = {
+            header_name = "api-key",
+            header_value = "azure-key",
+            allow_auth_override = false,
+          },
+          model = {
+            name = "gpt-3.5-turbo",
+            provider = "azure",
+            options = {
+              max_tokens = 256,
+              temperature = 1.0,
+              upstream_url = "http://"..helpers.mock_upstream_host..":"..MOCK_PORT.."/llm/v1/chat/good",
+              azure_instance = "001-kong-t",
+              azure_deployment_id = "gpt-3.5-custom",
+            },
+          },
+        },
+      }
       --
 
       -- 200 chat bad upstream response with one option
@@ -442,6 +472,105 @@ for _, strategy in helpers.all_strategies() do if strategy ~= "cassandra" then
         }, json.choices[1].message)
       end)
 
+      it("good request with client right auth", function()
+        local r = client:get("/azure/llm/v1/chat/good", {
+          headers = {
+            ["content-type"] = "application/json",
+            ["accept"] = "application/json",
+            ["api-key"] = "azure-key",
+          },
+          body = pl_file.read("spec/fixtures/ai-proxy/openai/llm-v1-chat/requests/good.json"),
+        })
+
+        -- validate that the request succeeded, response status 200
+        local body = assert.res_status(200 , r)
+        local json = cjson.decode(body)
+
+        -- check this is in the 'kong' response format
+        assert.equals(json.id, "chatcmpl-8T6YwgvjQVVnGbJ2w8hpOA17SeNy2")
+        assert.equals(json.model, "gpt-3.5-turbo-0613")
+        assert.equals(json.object, "chat.completion")
+
+        assert.is_table(json.choices)
+        assert.is_table(json.choices[1].message)
+        assert.same({
+          content = "The sum of 1 + 1 is 2.",
+          role = "assistant",
+        }, json.choices[1].message)
+      end)
+
+      it("good request with client wrong auth", function()
+        local r = client:get("/azure/llm/v1/chat/good", {
+          headers = {
+            ["content-type"] = "application/json",
+            ["accept"] = "application/json",
+            ["api-key"] = "wrong",
+          },
+          body = pl_file.read("spec/fixtures/ai-proxy/openai/llm-v1-chat/requests/good.json"),
+        })
+
+        local body = assert.res_status(401 , r)
+        local json = cjson.decode(body)
+
+        -- check this is in the 'kong' response format
+        assert.is_truthy(json.error)
+        assert.equals(json.error.code, "invalid_api_key")
+      end)
+
+      it("good request with client right auth and no allow_auth_override", function()
+        local r = client:get("/azure/llm/v1/chat/good-no-allow-override", {
+          headers = {
+            ["content-type"] = "application/json",
+            ["accept"] = "application/json",
+            ["api-key"] = "azure-key",
+          },
+          body = pl_file.read("spec/fixtures/ai-proxy/openai/llm-v1-chat/requests/good.json"),
+        })
+
+        -- validate that the request succeeded, response status 200
+        local body = assert.res_status(200 , r)
+        local json = cjson.decode(body)
+
+        -- check this is in the 'kong' response format
+        assert.equals(json.id, "chatcmpl-8T6YwgvjQVVnGbJ2w8hpOA17SeNy2")
+        assert.equals(json.model, "gpt-3.5-turbo-0613")
+        assert.equals(json.object, "chat.completion")
+
+        assert.is_table(json.choices)
+        assert.is_table(json.choices[1].message)
+        assert.same({
+          content = "The sum of 1 + 1 is 2.",
+          role = "assistant",
+        }, json.choices[1].message)
+      end)
+
+      it("good request with client wrong auth and no allow_auth_override", function()
+        local r = client:get("/azure/llm/v1/chat/good-no-allow-override", {
+          headers = {
+            ["content-type"] = "application/json",
+            ["accept"] = "application/json",
+            ["api-key"] = "wrong",
+          },
+          body = pl_file.read("spec/fixtures/ai-proxy/openai/llm-v1-chat/requests/good.json"),
+        })
+
+        -- validate that the request succeeded, response status 200
+        local body = assert.res_status(200 , r)
+        local json = cjson.decode(body)
+
+        -- check this is in the 'kong' response format
+        assert.equals(json.id, "chatcmpl-8T6YwgvjQVVnGbJ2w8hpOA17SeNy2")
+        assert.equals(json.model, "gpt-3.5-turbo-0613")
+        assert.equals(json.object, "chat.completion")
+
+        assert.is_table(json.choices)
+        assert.is_table(json.choices[1].message)
+        assert.same({
+          content = "The sum of 1 + 1 is 2.",
+          role = "assistant",
+        }, json.choices[1].message)
+      end)
+
       it("bad upstream response", function()
         local r = client:get("/azure/llm/v1/chat/bad_upstream_response", {
           headers = {