Could regulating the creators deliver trustworthy AI?

Labhaoise Ni Fhaolain, Andrew Hines

Abstract

Is a new regulated profession, such as Artificial Intelligence (AI) Architect who is responsible and accountable for AI outputs necessary to ensure trustworthy AI? AI is becoming all pervasive and is often deployed in everyday technologies, devices and services without our knowledge. There is heightened awareness of AI in recent years which has brought with it fear. This fear is compounded by the inability to point to a trustworthy source of AI, however even the term "trustworthy AI" itself is troublesome. Some consider trustworthy AI to be that which complies with relevant laws, while others point to the requirement to comply with ethics and standards (whether in addition to or in isolation of the law). This immediately raises questions of whose ethics and which standards should be applied and whether these are sufficient to produce trustworthy AI in any event.

Introduction: The Crisis of Trust in AI

As artificial intelligence (AI) becomes increasingly integrated into the fabric of daily life—often operating invisibly within consumer devices and essential services—the gap between technological capability and public trust continues to widen. Current efforts to secure "trustworthy AI" are often caught in a reactive cycle. This paper argues that we are currently trapped in a regulatory "cat-and-mouse" game where traditional oversight mechanisms fail to keep pace with innovation.

The authors identify three systemic failures in the current landscape:

  1. The Legislative Lag: The speed of AI development far outstrips the ability of legislatures and judiciaries to draft and implement laws. By the time a regulation is enacted, the underlying technology has often evolved beyond its scope.
  2. The Enforcement Gap: While ethical frameworks (such as corporate "don’t be evil" pledges) and industry standards (like those from the IEEE) exist, they frequently lack "teeth." They remain aspirational guidelines rather than enforceable mandates.
  3. The Jurisdictional Muddle: AI development is borderless, making it difficult for individual nation-states to impose effective top-down rules on global entities.

Methodology: Learning from Established Professions

To address these failures, this research synthesizes concepts from law, ethics, and the sociology of professions. The authors conducted a literature review of high-stakes fields—specifically medicine, law, and civil engineering—to understand how these sectors achieved and maintained public trust.

By analyzing existing frameworks like the GDPR and EPSRC principles, the paper concludes that regulating the product (the code) or the process (the development lifecycle) is insufficient. Instead, the authors advocate for a "bottom-up" approach: regulating the practitioner.

The Proposal: The "AI Architect"

The central thesis of this paper is the creation of a protected, regulated professional title: the AI Architect. Much like a Professional Engineer or a Medical Doctor, an AI Architect would be personally accountable for the systems they design and deploy.

Key Components of the Professional Model:

  • Personal Accountability: To market a product as "Trustworthy AI," a certified AI Architect would be required to sign off on the system. This places their professional license and right to practice on the line, moving accountability from an abstract corporate entity to a specific individual.
  • Regulating the Person, Not the Code: While software code is difficult to litigate due to its complexity and abstract nature, a professional is bound by a clear code of conduct. Facilitating unethical or illegal practices would result in disciplinary action or the revocation of the practitioner’s license.
  • The Professional Shield: One of the most significant insights is the potential to rebalance power between individual engineers and their employers. A regulated profession provides a "shield"; an engineer can refuse an unethical corporate request by citing their professional obligations to a regulatory body, mirroring how a doctor might refuse an unsafe medical procedure regardless of administrative pressure.
  • Trustworthy by Design: By embedding responsibility at the start of the development process, ethics become a core component of the lifecycle rather than a post-hoc compliance check.

Conclusion and Future Outlook

The research suggests that the technology industry has reached a tipping point where self-regulation is no longer viable. While a regulated profession is not a panacea, it offers a more agile solution than statutory law. Professional codes of conduct can be updated rapidly to address emerging technological risks, providing a flexibility that traditional legislation lacks.

However, several hurdles remain for future implementation. Defining the specific competencies and educational requirements for an "AI Architect" is a complex task, especially given the internal role-distribution conflicts within computer science. Furthermore, the borderless nature of AI continues to challenge state-level professional regulation. Despite these challenges, the paper concludes that personal responsibility is the missing link in the quest for truly trustworthy AI.

Summary updated

Abstract

Is a new regulated profession, such as Artificial Intelligence (AI) Architect who is responsible and accountable for AI outputs necessary to ensure trustworthy AI? AI is becoming all pervasive and is often deployed in everyday technologies, devices and services without our knowledge. There is heightened awareness of AI in recent years which has brought with it fear. This fear is compounded by the inability to point to a trustworthy source of AI, however even the term "trustworthy AI" itself is troublesome. Some consider trustworthy AI to be that which complies with relevant laws, while others point to the requirement to comply with ethics and standards (whether in addition to or in isolation of the law). This immediately raises questions of whose ethics and which standards should be applied and whether these are sufficient to produce trustworthy AI in any event.
BibTeX
@misc{fhaolain2020could,
  author = {Labhaoise Ni Fhaolain and Andrew Hines},
  title = {{Could regulating the creators deliver trustworthy AI?}},
  year = {2020},
  eprint = {2006.14750},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2006.14750},
}
Figure