Monitoring Misuse for Accountable 'Artificial Intelligence as a Service'

Abstract

AI is increasingly being offered 'as a service' (AIaaS). This entails service providers offering customers access to pre-built AI models and services, for tasks such as object recognition, text translation, text-to-voice conversion, and facial recognition, to name a few. The offerings enable customers to easily integrate a range of powerful AI-driven capabilities into their applications. Customers access these models through the provider's APIs, sending particular data to which models are applied, the results of which returned.However, there are many situations in which the use of AI can be problematic. AIaaS services typically represent generic functionality, available 'at a click'. Providers may therefore, for reasons of reputation or responsibility, seek to ensure that the AIaaS services they offer are being used by customers for 'appropriate' purposes.This paper introduces and explores the concept whereby AIaaS providers uncover situations of possible service misuse by their customers. Illustrated through topical examples, we consider the technical usage patterns that could signal situations warranting scrutiny, and raise some of the legal and technical challenges of monitoring for misuse. In all, by introducing this concept, we indicate a potential area for further inquiry from a range of perspectives.

Role of artificial intelligence in cloud computing, IoT and SDN: Reliability and scalability issues

“…Besides the services provided by the cloud to consumers, the use of artificial intelligence as an artificial intelligence-as-a-service (AIaaS) [21] is also available to the users. Various fields of AI like neural networks and machine learning, gather vast data sets to construct, train, process, and execute the models effectively.…”

Artificial Intelligence as a Service

“…Following the market trends of AIaaS, researchers in diverse disciplines, including information systems, computer science, and management, have started to focus their research on provisioning AI capabilities from the cloud. Diverse research streams on AIaaS recently emerged that deal, for example, with the design and evaluation of AI services (Boag et al 2018; e.g., Elshawi et al 2018), the adoption and effective use of AIaaS (e.g., Zapadka et al 2020;Pandl et al 2021), uncovering AIaaS misuse by its users (e.g., Javadi et al 2020), or understanding AIaaS's issues and vulnerabilities (e.g., Truex et al 2019).…”

“…The research field on AIaaS itself is still scattered and combines terminologies and approaches from multiple disciplines. While the term ''artificial intelligence as a service'' is seldom found in the literature (e.g., Javadi et al 2020;Zapadka et al 2020), researchers and practitioners use an ever-increasing amount of different terms to describe the phenomenon. ''Machine learning as a service'' is certainly most widely encountered in the literature (Duong and Sang 2018;e.g., Yao et al 2017), but related terms are also, such as ''deep learning as a service'' (e.g., Boag et al 2018), ''inference as a service'' (e.g., Romero et al 2019), ''neural networks as a service'' (Huqqani et al 2014), or ''analytics as a service'' (e.g., Naous et al 2017), among others.…”

“…These terms are mostly driven by practice, innovations, and the ever-increasing number of offerings on the market. In addition, these terms mostly cover AI software and applications, and thus AIaaS literature mostly relates to the conventional software as a service (SaaS) cloud model (e.g., Javadi et al 2020). On the contrary, cloud providers have already started offering AI developer services and AI infrastructure services, relating to the conventional platform (PaaS) and infrastructure as a service (IaaS) cloud models that have been neglected by prevalent research so far.…”

“…The most prominent and frequently used types of AIaaS are AI software services that are ready-to-use applications or building blocks (Javadi et al 2020). They relate to the conventional SaaS cloud models (cf.…”