Skip to main content

Appendix D | AIAT Landscape Scan Approach

This appendix details the approach employed to survey the commercial landscape in AI risk mitigation, as outlined in our Report. Landscape Scan Process:

Our research, conducted from January to April 2024, began with two parallel streams:

  1. A comprehensive literature review covering all concepts, pilots, tools, products, and services aimed at mitigating AI-related harms and risks.
  2. A scan and screening of early-stage, growth-stage, and established companies with AIAT-related offerings.

This process resulted in the discovery of 145 organisations (100 startups, 12 established enterprises, and 33 nonprofits or other types of entities), and lead to the classification of product and service ideas into the following four "solution domains": AI-Resilient IT Security (safeguarding), AI Trustworthiness (auditing), AI-Centric Risk Management (governing), and AI-Aware Digital Authenticity (verifying). Scope and Limitations:

  • It is important to note, our landscape scan provides a market sampling and is not exhaustive. The scope of our research did not afford a comprehensive examination of all existing entities within the AIAT market. Moreover, given the rapidly evolving nature of AI risk mitigation, some emerging companies might not have been accurately captured.

  • Our focus was primarily on English-speaking companies in Western countries, introducing a geographical bias that may have overlooked innovative solutions from non-English speaking regions.

Market Segmentation Challenges:

  • Companies were categorised based on their self-described offerings. Despite segmenting the market into discrete "solution domains" focused on safeguarding, auditing, governing, or verifying, the diverse nature of these companies often sees them spanning multiple domains. This is because many companies offer multiple product lines, each serving different functions. This complicates their classification into a single domain, highlighting a significant challenge in neatly segmenting this evolving market.

  • Given the inclusion of ideas from a broad literature review, some descriptions included in the "How might it work?" subsections are not summaries of existing marketable solutions, but rather predictions about prospective technologies that may be applied in commercial contexts. This approach aims to provide a forward-looking insight into potential future developments in AI risk mitigation.

Exclusions from the report:

  • Nonprofits, research organisations, and open-source frameworks were excluded from the company samplings in this Report. The Report focused on commercial companies offering third-party solutions to mitigate AI risks. Additionally, we have also excluded incumbent companies that are potentially well-resourced and able to pivot towards AIAT -related solution offerings. These established companies often involve more human capital -intensive services, and are unlikely to require the kind of funding that this report aims to support.