Share with your CIO
Malta has established a national AI Governance Council led by Chief AI Officer Jonathan Gerada, positioning the country as one of the few governments with a dedicated cross-agency AI coordination body. The Council’s operating principle is “analyse, simplify, automate,” meaning no process gets automated before someone asks whether it’s worth automating at all. Role-specific AI training for accountants, HR officers, and procurement professionals is being coordinated centrally, moving beyond the broad AI literacy subscription Malta already extended to every citizen.
What this means for your business
The CIOs most exposed to this story aren’t in Malta. They’re the ones running enterprise AI programs that look exactly like what Gerada is trying to avoid: scattered deployments across business units, governance that lives in a policy document rather than in operational decisions, and “lift and shift” automation that digitizes broken processes rather than fixing them. If your AI program is measured by the number of tools deployed rather than by measurable process outcomes, Malta’s framework is describing your problem, not your solution.
The governance model here deserves more credit than it typically gets in enterprise contexts. Most large organizations treat AI governance as a compliance layer added after deployment decisions are made. Malta is structuring it the opposite way, with the Council sitting upstream of deployment rather than auditing downstream of it. The “action-takers rather than paper-pushers” framing is doing real work: it signals that the Council has operational authority, not just advisory status. Enterprises that have built AI review boards without giving them budget authority or the power to block projects will recognize the gap immediately.
The accountability argument Gerada makes cuts sharper than it sounds. “You cannot have accountability on a machine” is not a philosophical position; it’s an organizational design constraint. Every AI deployment that lacks a named human owner with genuine decision authority creates a liability pocket, one that regulators and auditors are becoming increasingly skilled at finding. The CIOs who will be ahead of this aren’t the ones waiting for regulation to force the issue. They’re the ones who can already name the person accountable for each deployed model. If that answer is “the vendor,” the governance structure needs rebuilding before the audit does it for them.
Concept deep-dive: Human-in-the-loop governance
Human-in-the-loop governance means a person with actual authority reviews, approves, or can override an AI system’s output before it produces a consequential result, rather than simply monitoring outcomes after the fact. Think of it as the difference between a co-pilot who can take the controls versus a flight recorder that captures the crash. In enterprise terms, it determines who owns the decision when the model is wrong, which is the question every regulator and plaintiff’s attorney will eventually ask.
Based on reporting from Malta forms AI Governance Council to put people at the centre of AI adoption, originally published 2026-08-28 19:03:00.
