Who Governs the Governors? Transparency Must Apply to AI Oversight Too
The conversation around AI governance took another important turn this week.
On August 3, the White House finalized a voluntary framework for evaluating the cybersecurity risks of frontier AI models, bringing together leading AI developers—including OpenAI, Anthropic, Google, and Meta—to participate in the process.
At first glance, this is exactly the kind of initiative many have been calling for. As AI systems become increasingly capable, governments need mechanisms to evaluate potential risks before deployment. Independent assessment, structured testing, and collaboration between regulators and developers are all positive steps toward responsible AI governance.
But one important question has received far less attention.
Who governs the governors?
Reports indicate that key benchmarks, evaluation thresholds, and testing methodologies used within the framework will not be made public. While some degree of confidentiality may be justified for national security or to prevent misuse, this also means that neither researchers, industry, nor the public can independently evaluate how rigorous—or consistent—these assessments actually are.
Transparency has become one of the defining principles of AI governance.
Only days earlier, the European Union began enforcing the transparency provisions of the AI Act, requiring organizations to disclose AI interactions, identify AI-generated content, and document how AI systems are used.
These requirements reflect an important principle:
Trust is built through transparency.
Yet transparency should not apply only to organizations deploying AI. It should also apply—wherever reasonably possible—to the institutions responsible for overseeing AI.
This does not mean governments must publish every technical detail or disclose information that could compromise national security. But meaningful governance requires more than asking the public to trust that robust evaluations have taken place.
It requires confidence that oversight itself follows accountable, well-defined, and defensible processes.
If the benchmarks, thresholds, and evaluation methods remain undisclosed, an important question remains unanswered:
How can society evaluate the quality of AI oversight if the oversight process itself cannot be independently examined?
At Human Continuity Organization, we believe that accountability extends throughout the entire AI ecosystem.
Organizations should be transparent about how they deploy AI.
Developers should be transparent about how their systems operate.
And institutions responsible for evaluating AI should be transparent about the principles, governance structures, and decision-making processes that underpin their assessments.
Transparency is not simply an obligation placed on organizations—it is the foundation upon which public trust in AI governance is built.
As governments around the world continue developing AI oversight frameworks, this principle deserves to remain at the center of the conversation.
Because effective AI governance is not only about regulating technology.
It is also about ensuring that those who regulate are themselves accountable.