Algoithm & Blues Vol. 14: Can AI Govern AI? Exams Say…Maybe.

Researchers recently gave frontier LLMs the same certification exams that human privacy and governance professionals sit for—CIPP/US, CIPT, CIPM, even the new AIGP. On paper, the results look impressive: most models scored at or above passing levels. In theory, an AI could walk away with credentials that many compliance officers build careers around.

But there’s a catch. The weakest performance came on the CIPM—the program management exam. The models handled statutes, definitions, and regulatory frameworks with ease. The messy, context-heavy work of running a governance program—managing people, processes, and accountability? That’s where they couldn’t quite pull it off.

This reinforces a critical point: AI is excellent at structured, well defined knowledge work but much less capable when success depends on navigating organizations, personalities, and incentives. It can draft a privacy policy, cite precedents, and check compliance clauses at speed. But deciding how that policy fits into an enterprise’s workflows, or negotiating among legal, IT, and business stakeholders—that still takes people.

Consider a practical scenario: a bank is updating its incident response plan for AI-driven fraud detection. An LLM can:

• Summarize the latest regulatory guidance in plain language.

• Cross-check the draft plan against known standards.

• Generate a comparative table of obligations across jurisdictions.

What it can’t do is broker agreement between legal, risk, and operations on who owns each escalation step, or convince the board that the plan is credible. That’s human work—judgment, trust, and authority.

The right framing isn’t AI stepping into the role of chief compliance officer. It’s more like AI as the research clerk—fast, tireless, and precise—while people still hold the gavel. Use it to accelerate the rote and the reference-heavy, while keeping accountability, coordination, and the final signature firmly human.

The Takeaway:

Boards should expect faster first drafts of policy, sharper monitoring, and quicker cross-checks—but regulators will still demand human accountability. Governance is about alignment as much as it is about law, and that part isn’t automatable.

AI may be able to master the rules, but it still struggles with the org chart.

hashtag #AI hashtag #Governance hashtag #Compliance hashtag #RiskManagement hashtag #FutureOfWork

https://lnkd.in/eUHC6Zjk …more