AI Auditing is a crucial process in the development and deployment of AI, ensuring that these systems operate in a manner that is transparent, unbiased, and accountable. As AI models become more complex and their adoption becomes more widespread across industries, there is increased need for auditing these systems to evaluate their operations, identify potential biases or flaws, and mitigate any risks that could arise from their misuse. AI Auditing may involve aspects such as model explainability, input transparency, ethical oversight, and system robustness. It's an essential part of building trust in AI systems, and maintaining regulatory compliance.
The term "AI Auditing" emerged in the mid-2010s, as the implications of widespread AI use began to catch the attention of ethicists, regulators, and the public. The need for auditing has grown in tandem with the increasing adoption of AI across different sectors, and the realization of different types of complications and ethical issues that can arise from automated decision making.
Given the interdisciplinary nature of AI auditing, there isn't singular contributors. Instead, a range of professionals from fields such as data science, ethics, law, and policy have been critical in shaping the practice and discussion of AI auditing. Organizations such as Algorithm Watch, the AI Now Institute, and Fairness, Accountability, and Transparency in Machine Learning (FATML) have played a significant role in advocating for and developing practices related to auditing AI systems.