The Artificial Intelligence (AI) global market is growing,  and its growing fast.  For the past five years we have witnessed high revenue growth for the industry, year-in year-out, which is the reason for the shorter innovation-to-product cycles we see in the AI sector.  It is no surprise then, to see increasing AI adoption in society due to these innovation-speeds and the fact that the technology is easily adaptable for day-to-day tasks.

Every week new versions of existing AI products are announced with the industry witnessing the arrival of a 'new kid on the block', from time-to-time.  Every week the technology is producing faster, better and solid AI products that seem to out-perform the expectations of even the experts in the field.  Well researched academic papers are being 'dropped' at every turn capturing ideas for new and improved technologies that can enhance AI machine capabilities.

One cannot help but imagine that law-makers, by now, are experiencing some form of 'paralysis' in triangulating what needs to be done to keep-up with these fast-paced developments.  You could imagine further, that the 'paralysis' is exacerbated by the technical knowledge required to craft effective AI regulations/policies.

On the side, jurisdictional challenges inherent in the centralized AI-talent and AI-infrastructure means that in most regions (like Africa) where there is lack of control of AI economics,  'paralisis' in AI governance and regulation is unavoidable. Simply put, if your region has little or no control of AI talent and infrastructure, even with robust regulations & policies your impact is minimal - having to conform to "EthicalAI" as defined elsewhere.

Context: Why Regulation is Urgent

A key concern for today's law-makers is the question of what regulation & measures should be put in place in order to align AI implementation with well-defined principles, public norms and values that would prevent 'the Machines' from causing harm to human beings.

The understanding is that the Machines are devices that need to be coded for all environments in which they operate.  That, at this juncture they do not possess the ability to discriminate between 'wrong' or 'right', which poses a risk when engaging with human beings. 

Artificial Intelligence (AI) is now part of everyday life  —  from job applications and banking to healthcare and education.  The time is now for its regulation so as to ensure the protection of the human species going forward.

A case in point, is one in which Amazon had implemented an 'intelligent-tool' trained on processing résumés/cv's submitted by applicants for technical roles within the company - a tool that was used for almost ten-years.

 The tool was trained primarily on male résumés which resulted in the tool favoring male applicants despite not having been explicitly programmed to do so.  On closer investigation, it was found that the AI-tool penalized résumés containing the word “women’s” (e.g., “women’s only college”), reinforcing existing gender disparities in hiring.

This case illustrates how historical biases embedded in training data can lead AI models to unintentionally discriminate and perpetuating systemic inequalities. The tool was finally scrapped by Amazon for being biased against female applicants.

Governments and policy-makers around the world are asking:

    "How do we make sure AI is fair, safe, and doesn’t harm people?"

Principles and Approaches to Achieving Ethical AI

Ethical AI refers to the development and deployment of artificial intelligence systems that uphold fairness, transparency, accountability, and respect for human values.  Four key concerns shape the ethical discourse around AI:

• Bias:   AI systems risk reinforcing discrimination due to biased training data or algorithmic design.
• Explainability:  AI decisions must be interpretable and understandable to ensure transparency and trust.
• Robustness:   AI should be resilient against adversarial attacks, errors, and unintended consequences.
• Privacy:  AI must protect personal data and prevent security breaches or misuse of sensitive information.

To mitigate these risks the following measures are neccessary:

• Embedding Risk Mitigation into AI Systems:
  Incorporating both technical and non-technical safeguards to prevent ethical breaches.
• Enhancing Accountability and Transparency:
  Requiring explain-ability in algorithmic decision-making and system design.
• Developing Ethical Decision-Making Frameworks:
  Setting standards for AI model training, ethical system design, and public education.
• Implementing Regulatory Oversight:
  Establishing oversight bodies at different levels to monitor AI development and deployment.
• Promoting a Culture of Responsible AI:
  Encouraging AI developers and deployers to adopt ethical best practices and uphold high moral standards.

We have seen the U.S and China putting in place regulations and mechanisms that are fully capable of delivering the regulations and mechanism that implement these requirements, no doubt.

As stated before, the challenge is not in the lack of a capacity to regulate (in many instances) but the challenge is in the fast-paced iterating-versions that hit the 'innovative-run-away' each and every week, talent and ownership and jurisdictional issues.

Hence, regulators are called upon to be vigilant and guard against these risks for the sake of the general public's interest, health and protection.

Timing, Expertise & Jurisdictional Issues

AI systems may be created by code, but their impacts are profoundly human.   Effective regulation of AI to prevent discrimination is not just about drawing up rules  — it’s about redefining the social contract in the digital age,  where justice,  fairness,  and human dignity must be embedded directly into the logic of the machines.

The issues with regards to aligning AI to societal standards like fairness and equity revolve around the practical operationalization methods employed prior the event of implementation.  For effective regulation of AI,  I would say, three consideration are critical:

• efforts to reduce the adverse impact of the technology by attending to speed of regulating, 
• the acquisition of technical expertise around the technology itself,  and
• the implementation of strategies to overcome jurisdictional limitations.

Timing: Regulation vs. Innovation Speed

The rapid evolution of AI far outpaces the slow cycles of policy-making, creating a significant gap between technological advancements and regulatory responses.   While laws can take months or even years to enact, AI systems can be updated,  retrained, or deployed almost instantly.

This disparity poses serious challenges, as many existing laws are becoming obsolete,  having been crafted long before the emergence of today’s advanced AI tools.   For instance, harmful applications of AI  —  such as deepfakes, violations of personal privacy,  or the use of AI in autonomous weapons —  were beyond the imagination of lawmakers when those laws were written.

Experts in Technology Governance argue that expecting regulators to keep pace with the speed of AI innovation is unrealistic.    Instead, they propose for the implementation of relevant strategies aimed at bridging this gap without overburdening regulators and policymakers.

These approaches are designed to address the unique challenges posed by AI's rapid development while ensuring that governance frameworks remain effective and relevant.  To bridge this gap,  we need adaptive, proactive, and multi-stakeholder approaches:

1. Agile & Principles-Based Regulation

That regulators should design "Outcome-Focused Rules" and move away from rigid and technical pronouncements.  That specifications of their laws should define risk-based thresholds (e.g., "AI must not discriminate") that remain relevant even as tech evolves.

Use "Sand-boxing & Experimentation" techniques which are a mechanism that allows controlled real-world testing of AI under temporary regulatory waivers (as seen in the UK’s Financial Conduct Authority sandbox).  Lastly, this approach calls for the use of "Sunset Clauses" in regulating this industry   -  which are laws that auto-expire after a set period, forcing regular updates.

2. Industry Self-Regulation [with Teeth]

That governments should encourage the formation of "Dynamic Standards Bodies" which are voluntary and industry based consortiums to update technical standards in near-real time.  In the same spirit "Certification & Auditing" requirements should be put in place in a way that allows for third-party audits for AI, with penalties for non-compliance.

3. Public Empowerment & Decentralized Oversight

There is a strong public call for "Citizen Review Boards" which establish crowd-sourced oversight of AI systems (e.g.,  Mozilla’s "Let’s Audit AI" initiative).  Again, "Whistle-blower Protections regulation might be a need in order to shield tech insiders who expose unethical AI practices - perhaps as envisaged in the EU’s whistle-blower directive.  Also, strengthening the legal machinery through re-skilling in AI based technologies is a program to be considered by regulators, inclusive of bringing to bear such training with the "Fourth-Estate".

---

Thanks for reading! For exclusive insights and to help keep this blog thriving, join our Patreon community today!