Governance in nascent AI ecosystems should focus on developing strategies that create an enabling environment for innovation, setting the stage for a robust industry and, eventually, effective regulation.
On June 2024, the Public Petitions Committee of the National Assembly, chaired by Earnest Kagesi, met with members affiliated to the Robotics Society of Kenya in connection with a public petition on the Kenya Robotics and Artificial Intelligence Society Draft Bill, 2023.
The Bill seeks to establish the Kenya Robotics and Artificial Intelligence Society, which would be responsible for regulation, licensing, strategy formulation, awareness creation, and capacity building.
While we share the goal of advancing the robotics and AI sector and agree on the need for responsible technologies, concur with the guiding principles and concede that regulation will eventually be necessary; hurried regulatory attempts such as these, predicated on Global North regulatory developments, are not only misadvised - based on timing and mandate - but also detrimental to the growth of nascent robotic and AI ecosystems situated in developing regions.
One key function of the proposed society is to regulate the robotics and AI sector, assuming its robustness in Kenya. Although significant activity is taking place within the sector, as evidenced by considerable start-up growth, AI convenings and recent infrastructural investments, this does not equate to a robust industry. Such activity within a nascent AI ecosystem should not be conflated with a sector.
Kenya currently ranks 101 in the "Government AI Readiness Index 2023" report. Simply put, there is no robotics and AI industry to talk of in the country.
To foster the growth of such an industry, which eventually would necessitate regulation, immediate action is needed to address key challenges: the dearth of datasets, exorbitant compute costs, limited infrastructure (electricity, connectivity, devices), low levels of local AI talent and governance.
Establishing the society at this point is not only premature but detrimental to the ecosystem as it detracts from the real issues facing the AI ecosystem, chokes innovation, discourages investors and limits the transformative impact of the sector on the economy.
The proposed society's mandate of regulation, awareness creation and capacity building, is beyond its scope and duplicates existing efforts within the sector. Such a mandate demands multi-stakeholder collaboration, a Herculean task for the proposed society.
Numerous stakeholders, including ministries, private players, educational institutions and non-governmental organisations are currently involved in undertaking the proposed mandate of the society. Insisting on this duplicative mandate adds an unnecessary burden on taxpayers and creates another bureaucratic hurdle for practitioners revealing an egotistical pursuit rather than a practical solution.
Regarding the way forward, we outline four propositions to inform the regulation discourse not only in Kenya but also in other nascent AI ecosystems. First, pause on regulatory attempts, which in our context means the cessation of discussions on a petition regarding the draft Bill.
Second, leverage existing legislation, institutions and initiatives to inform a thoughtful approach to robotics and AI governance. AI technologies, while novel in their impact and speed of development, remain simple in form: data is collected, AI models are developed, and then deployed to potential users.
Embed guiding principles early in the data collection process by utilising ethics boards across the country. This ensures that projects align with the Data Protection Act No 24 of 2019 and allows domain experts from the various spheres that AI will influence to flag issues at the conception stage.
The recently operationalised Regulatory Sandbox Framework, by the Communications Authority of Kenya (CA), is an apt provision to spur AI development and is an alternative regulatory tool that allows innovators to test emerging ICT products and services in a controlled environment. It encourages innovation by providing a safe space to address potential regulatory issues during development.
Third, rely on existing consumer protection laws and institutions such as the Consumer Protection Act of 2012, Ministry of Information, Communications and The Digital Economy, Office of the Data Protection Commissioner, and the CA to address potential issues during deployment. Where current mechanisms are insufficient, amendments can be considered.
Fourth, leverage current initiatives by the government such as the Working Group on Policy and Legislative Reforms for the Information inform a thoughtful approach to AI governance.
Such an avenue provides for a comprehensive and inclusive engagement of stakeholders within the AI ecosystem, enabling the integration of diverse perspectives to inform the design of a national strategy on Robotics and AI, financing of research and development, fostering of public-private partnerships, upgrading of AI infrastructure and the responsible use of Robotics and AI-based technologies.
These recommendations are just but a tip of the solution; they pave the way toward creating a formidable industry that can merit regulation. Regulation must follow strategy. Governance in nascent AI ecosystems should focus on developing strategies that create an enabling environment for innovation, setting the stage for a robust industry and, eventually, effective regulation.
Antony Ndolo, Kavengi Kitonga, Lynda Ouma and Paul Amayo are researchers.
