AutoML Model says response headers are too long during real-time inference

I trained a binary classification model using SageMaker AutoML, and deployed an endpoint for real-time inference via the UI (i.e. clicked "Deploy Model" after AutoML made the model).

The JSON I send for inference is just text, we aren't doing images or anything.

When I am in the "Endpoint Details" and send the test data for inference, it says: Error invoking endpoint: Received server error (0) from primary with message "Response received from the primary container has headers with length greater than 4096 bytes. Reduce the length of your container's response headers and update your endpoint.". See https://us-east-2.console.aws.amazon.com/cloudwatch/home?region=us-east-2#logEventViewer:group=/aws/sagemaker/Endpoints/XXXXXXX in account XXXXXX for more information.

The Access Log the error message links to just says: 2023-11-15T03:04:05,186 [INFO ] W-9000-model ACCESS_LOG - /XXXXXX "POST /invocations HTTP/1.1" 500 14 2023-11-15T03:04:05,186 [INFO ] W-9000-model ACCESS_LOG - /XXXXXX "POST /invocations HTTP/1.1" 500 14

Having made the model with AutoML, rather than coding it myself, I have no idea how to modify the response headers that the model creates.

How do I fix this?