A reported agreement requires the tech giant to help adjust its AI safety settings and filters at the government's request.