Artificial intelligence is moving from experimental chatbots into workplaces, schools, businesses and government services, but a new question is emerging globally: Who checks whether powerful AI systems are safe and accurate?
California has taken one of the clearest steps yet towards answering that question, introducing a framework for independent third-party assessments of AI systems.
On September 9, California Governor Gavin Newsom signed two laws establishing a framework for independent organisations to assess AI systems and creating a registry for AI auditors.
The measures are designed to ensure that developers do not have to rely solely on their own internal assessments when determining whether their AI systems meet required standards.
The move comes as the capabilities of AI systems continue to expand.
AI tools are increasingly being used to generate text and images, analyse information, write computer code, automate customer service and assist with decision-making.
But the technology can also produce inaccurate information, expose sensitive data or behave unpredictably in certain circumstances.
That has prompted a growing debate over whether companies developing the most powerful systems should be responsible for assessing their own risks.
The issue is particularly relevant to Kenya, where AI adoption is also accelerating.
Parliament recently called for stronger oversight as government agencies increasingly turn to AI to improve service delivery, with lawmakers warning that technological development needs to be accompanied by accountability and safeguards for citizens.
The Kenyan debate therefore mirrors a much larger global question: how should governments regulate a technology that is developing faster than many existing laws?
Independent auditing is one possible approach.
Instead of asking an AI company to simply declare that its system is safe, external experts can examine how the system works, test it for risks and assess whether claims made by its developer can be independently verified.
California’s new framework establishes standards concerning the independence, transparency and integrity of AI auditors.
The debate has also intensified within the technology industry itself.
Anthropic chief executive Dario Amodei recently called for the development of advanced AI to be paced more carefully, warning about potential risks from increasingly capable AI systems and AI agents.
Meanwhile, Microsoft co-founder Bill Gates has warned that governments globally are not sufficiently prepared for the changes AI could bring, including disruption to jobs and cybersecurity risks. At the same time, he has argued that AI could deliver significant benefits in areas such as healthcare, agriculture and education if access is managed fairly.
For Kenya, the debate presents both a regulatory and economic challenge.
AI could help businesses automate routine work, help farmers analyse information, support doctors and expand access to educational tools.
But if AI systems are used to make decisions affecting citizens, questions of accuracy, transparency, privacy and accountability become increasingly important.
For example, if an automated system assists with determining eligibility for a government service, assessing a loan application or filtering job applicants, citizens may reasonably want to know how the decision was reached and who is responsible when the system gets it wrong.
That is why the global discussion is shifting from simply asking what AI can do to asking how AI should be tested before it is trusted.
California’s approach does not automatically provide a model Kenya must adopt. Its laws apply within its own jurisdiction and regulatory system.
But the development points to a broader trend that Kenyan policymakers, technology companies and consumers will increasingly have to confront: as AI becomes more powerful, independent verification may become as important as the technology itself.
