Skip to main content

Stakeholder overview

There are a large number of parties that may be affected by, or involved in, the development, deployment and use of AI systems or AI system components and hence who might benefit from TAIBOM. They include the following:

  • Users of AI systems. These can be members of the public, governments or businesses that use products and services that involve AI. The presence of TAIBOM attestations for products and services can provide information relating to terms of use, security, privacy and performance.

  • Regulators. Regulators need to have confidence that AI systems in different contexts comply with regulatory requirements. The transparency provided by TAIBOM attestations should help regulators to do that.

  • Copyright holders and content owners. By providing transparency about the data used to train different systems, TAIBOM makes it easier for copyright holders to check that their content is being used appropriately.

  • Corporate legal teams. TAIBOM can make it easier to see which licences apply to a system that uses AI, making it easier for corporate legal teams to understand the resulting legal exposure.

  • Cybersecurity and IT security teams. TAIBOM makes it possible to see which emerging AI related security risks and vulnerabilities, such as prompt hijack attacks or data poisoning attacks, can be used against a business's AI assets.

  • AI system developers. Teams developing solutions using AI systems should more easily be able to understand the strengths and weaknesses of different AI systems, including independent benchmarks and known vulnerabilities.

  • Deployment teams. Teams deploying systems need to make sure that the inference system that is deployed into the production environment is the correct one and that the production environment is setup correctly.

  • Investors and venture capitalists. TAIBOM can support due diligence prior to an investment decision being made, making it easier to identify legal, technical, intellectual property and security risks arising from a business's AI assets.

  • AI ethics teams, academics and political actors. By improving transparency around the properties of high profile AI systems, TAIBOM can support the work of ethics teams, academics and political actors.

  • Insurance providers. Businesses that provide insurance against AI-related risks, such as copyright infringement, regulatory breaches, etc. will benefit from the trsanparency that TAIBOM provides about the properties of the systems being insured.

The stakeholders also include the following interested parties:

  • Hardware developers and vendors. TAIBOM should be able to represent information about training and inference hardware and firmware.

  • Software developers, and system and model builders, developers and vendors. TAIBOM should include information about the software components of an AI system, its training data and its training environment.

  • Service providers. TAIBOM should include information about inference software, inference data and the inference environment.

  • Providers of AI component repositories and AI component hosts and distributors. Providers of AI system or component stores or repositories should make TAIBOMs available for each AI system or component that they provide. Ideally, they should also allow user to search for AI systems and components based on the contents of their TAIBOMs.

Although all of these disparate stakeholders can produce and consume TAIBOM claims, it is primarily the AI end user that consumes the claims. And the primary purpose of this claim consumption is to evaluate the risk currently posed by the system.