The headline sounds ambitious
OpenAI and Malta announced a partnership this week to "bring ChatGPT Plus to all Maltese citizens." The framing is nation-scale, the rhetoric is sweeping, and George Osborne—OpenAI's Head of Countries—calls it a model for how "intelligence is becoming a national utility."
But when you read past the press release, the actual program looks a lot more conditional than the headline suggests. And that gap between messaging and mechanics matters, because this isn't just about Malta. OpenAI is positioning this as a template for other governments, and the details reveal some uncomfortable truths about what "AI for all" really means in 2025.
What's actually being offered
Here's the structure of Malta's "AI for All" initiative:
- First, citizens complete an AI literacy course developed by the University of Malta
- After completing the course, they become eligible for one year of free ChatGPT Plus
- The Malta Digital Innovation Authority manages distribution to "eligible participants"
- Phase one launches in May, scaling as more residents complete the course
This is not a universal rollout. It's a gated program where course completion is the price of admission. That's a defensible design choice—you could argue that pairing access with education is responsible. But it means the "all citizens" framing is doing a lot of rhetorical work that the actual mechanics don't support.
The course-completion gate raises questions
Requiring course completion before access creates a natural filter. Some citizens won't have time. Others won't see the value proposition. Still others will drop out partway through.
Malta has a population of around 520,000. The announcement doesn't specify how long the course takes, what the pass requirements are, or what "eligible participants" means beyond course completion. The phrase "scale as more Maltese residents and citizens abroad complete the course" suggests this is explicitly designed as a gradual rollout, not a Day One entitlement.
That's fine as a pilot program. But it makes the "world's first partnership to roll out ChatGPT Plus to all Maltese citizens" claim feel like overselling. You're not rolling it out to all citizens—you're offering it to citizens who complete a prerequisite course, managed through a government authority, on an undefined timeline.
The economics are deliberately opaque
The announcement doesn't mention cost. ChatGPT Plus is $20/month retail. If even half of Malta's population claimed a year of access, that's $520 million in retail value.
Obviously, OpenAI isn't charging Malta retail rates. The actual deal structure isn't disclosed, but the possibilities matter:
- Is Malta paying a per-user fee? A flat licensing deal? Nothing at all?
- Is this a marketing expense for OpenAI, effectively buying credibility with a small nation to establish the "Countries" template?
- What happens after the first year? Do users convert to paid subscriptions, does Malta renegotiate, or does access expire?
The lack of financial transparency makes it hard to evaluate whether this is a sustainable model or an expensive proof-of-concept that only works because Malta is small enough to subsidize.
What "AI literacy" actually means
The course is described as teaching "what AI is, what it can and can't do, and how to use it responsibly at home and work." That's reasonable on paper. But AI literacy courses vary wildly in quality, and this one is being built by the University of Malta with no public curriculum details.
The cynic in me wonders: is this a rigorous course on understanding model limitations, bias, and failure modes? Or is it a cheerful onboarding flow designed to maximize conversion to ChatGPT Plus usage?
The announcement positions the course as empowerment. But it's also a liability shield. If citizens misuse ChatGPT after completing a government-endorsed AI literacy course, Malta can point to the education requirement. The course isn't just pedagogy—it's political and legal cover.
The "intelligence as utility" framing is doing a lot of work
OpenAI's blog post opens with this: "We believe that, like electricity, intelligence should be available for people, businesses, and institutions to use as much as they need, where and when they need it."
It's a seductive metaphor, but it glosses over some key differences:
- Electricity is fungible and its output is predictable. LLMs are probabilistic and their failure modes are subtle
- Electricity infrastructure is owned and regulated as a public good. ChatGPT is a proprietary service controlled by a for-profit company
- Electricity doesn't hallucinate, introduce bias, or leak training data
Calling intelligence a "national utility" implies it should be treated like water or power—essential, regulated, and publicly accountable. But this partnership does the opposite: it outsources a core national capability to a private vendor with no disclosed source code, training data, or long-term pricing commitment.
What Malta actually gets right
Despite my skepticism, there are genuinely smart things about this program:
-
Pairing access with education is the right instinct. Too many AI adoption initiatives are just "here's the tool, good luck." Requiring literacy training—if it's done well—addresses the real problem that most people don't know how to use LLMs effectively.
-
Starting small with a pilot makes sense. Malta is small enough to iterate quickly. If this works, the lessons are portable. If it fails, the blast radius is contained.
-
The "OpenAI for Countries" model is more thoughtful than one-size-fits-all licensing. The announcement mentions that OpenAI is already working with Estonia and Greece on education-focused partnerships. Tailoring to national priorities is better than a generic enterprise contract.
The real question is what happens next
If this succeeds—let's say 200,000 Maltese citizens complete the course and actively use ChatGPT Plus—what does success actually look like?
- Do they become dependent on a tool they can't audit or replicate?
- Does Malta develop local AI expertise, or does it offshore cognitive infrastructure to San Francisco?
- What happens when GPT-5 or GPT-6 launches and Malta's deal covers only the current model tier?
- If OpenAI's corporate strategy shifts, does Malta have any leverage?
The optimistic read is that this partnership gives Malta a head start on AI adoption, training a generation of citizens to use LLMs fluently. The pessimistic read is that it's a subsidy that locks a small nation into dependency on a proprietary platform with no exit strategy.
Where this gets interesting for other countries
Malta is a test case. If OpenAI can make this work—and "work" means both meaningful adoption and positive press—they'll pitch the template to larger countries.
But the economics and politics get messier at scale. France, Germany, or Japan have populations 100–200x larger than Malta. The cost, the infrastructure load, and the political risk of outsourcing cognitive tools to a foreign vendor all increase non-linearly.
Malta's small size is a feature, not a bug. It lets OpenAI demonstrate the "Countries" model without the complexity of a large-scale rollout. But whether this is actually replicable at national scale is an open question.
The bottom line
This is a clever pilot program dressed up as a national transformation. The vision is compelling—who wouldn't want every citizen to have access to frontier AI?—but the execution is narrower and more conditional than the announcement suggests.
Malta is taking a real bet here, and I'm genuinely curious to see how it plays out. But let's be clear-eyed about what's actually being announced: a gated, course-conditional, time-limited program with undisclosed economics and no public accountability mechanism.
That's not a criticism of Malta—small countries have to take asymmetric bets to stay competitive. But it's a reminder that "AI for all" is a much harder problem than a press release makes it sound. Access is the easy part. Sustainable adoption, local capability-building, and avoiding vendor lock-in are the hard parts.
Where Malta leads, others may follow. But they should read the fine print first.