Northern Ireland’s government has been forced to write an artificial intelligence (AI) strategy without knowing which laws will be applied to businesses based there.
As the only part of the UK sharing a land border with an EU member state, it is caught between potential regulations from London and Brussels respectively.
Faced with a rulebook the Northern Ireland assembly cannot choose, the government’s approach has been to design governance that will hold up whichever rules are applied.
A few months ago, I reviewed Scotland’s AI Strategy.
What stood out was its positioning – by diverging from Westminster and aligning itself with the EU, it made a case for devolved governance on tech and digital regulation.
Northern Ireland now faces a similar constraint: how to balance between national UK legislation and a differing EU approach.
Unlike Scotland, this alignment choice is not entirely within the government’s control.
Straddling UK and EU AI Policy
The groundwork for the strategy in Northern Ireland has been laid for some time, and Belfast is aware of the stakes.
First Minister Michelle O’Neill was admirably blunt when launching the draft for public consultation.
Used well, AI can improve outcomes and make services more efficient, she said. But used poorly, “it risks undermining trust or creating unintended harms.”
What makes Northern Ireland’s AI position especially tricky is its physical borders.
As the only part of the UK connected to an EU member state by land, the arrangements that keep that border open with the Republic of Ireland are also those that keep a body of EU law in force in Northern Ireland.
When Brussels legislates, Northern Ireland can find itself inside the scope of rules that stop at the Irish Sea for everyone else in the UK.
The EU AI Act is testament to this. A handful of its provisions already apply in Northern Ireland, but the UK Government’s own assessment is that these have had little practical effect.
As the full rollout of the Act arrives, its risk classifications and corresponding obligations will be a different matter.
Northern Ireland could find itself subject to two (sometimes contradictory) sets of rules if the EU AI Act is fully enforced there, an approach that the European Commission has formally proposed.
As a result of Brexit negotiations, the Northern Irish Assembly can object to amended or replaced EU laws via the Stormont Brake.
But as the AI Act is new legislation rather than a revision, this “brake” cannot be “pulled.”
This means that the decision about Northern Ireland’s AI future will be made by the EU and UK Joint Committee, not by Northern Ireland itself.
What is striking is how little of this framing appears in the strategy itself.
The document runs to more than forty pages and mentions the EU AI Act exactly once, in the governance chapter, where it notes that the Republic of Ireland is implementing EU standards south of the border.
For a strategy that describes itself as “proudly place-based,” this is a curious silence about the single feature of the “place” which complicates legislation.
The practical implications are already clear.
Sixty miles south of Belfast, Ireland’s Regulation of AI Act 2026 is now law and an AI Office has been established in Dublin.
About 7,500 Northern Irish businesses trade goods and services worth roughly £10.5 billion (S$17.9 billion) across the border each year, and these companies will now face regulatory uncertainty.
Around nine in ten of these are SMEs.
If Northern Ireland ends up following UK rules alone, these firms will get a lighter domestic regime, but will still have to meet EU obligations to sell into the Republic because of EU AI Act obligations.
If Northern Ireland ends up inside the EU AI Act, these firms get a single rulebook aligned with their nearest and largest market but will carry heavier compliance costs than competitors in Britain.
While nothing is decided, these firms are bearing the worst version of both outcomes, preparing for two regimes at once or gambling on neither.
Fighting uncertainty with flexibility
Faced with this legislative no man’s land, the draft strategy concentrates almost entirely on the levers that the Executive holds.
If you cannot know what the rules will be, you must build for maximum compliance and maximum flexibility.
Inventories of AI use, human oversight teams, clear accountability routes and a workforce trained to challenge outputs are worth building, whichever way the legislation goes.
Building these systems would make Northern Ireland compliant under the EU AI Act and still credible under a lighter touch UK approach.
Failing to build them would be a misstep whatever the outcome.
The document is explicit that this is the approach.
It “does not prescribe technologies, platforms, or digital solutions for individual organisations,” but instead sets a direction while “providing the flexibility for each organisation to adopt the tools, approaches and recommendations that best suit their needs.”
It leads with principles, rather than rules, as prescribing specific requirements would mean guessing at the direction of legislation.
Eight principles therefore sit at the centre of the document: Human Oversight; Accountability and Redress; Data Governance; Technical Safety and Security; Fairness and Transparency; Sustainability; Societal Benefit; and Training and Literacy.
The strategy positions them against both possible futures quite deliberately, noting that the UK has five regulatory principles, whilst the EU has seven, and describing Northern Ireland’s eight as “future-facing.”
Read against the EU AI Act, they are recognisably its vocabulary. Read against the UK’s pro-innovation framework, they are unobjectionable.
This is a set of principles clearly built to be defensible under either legislative path.
Bridging the gap with low public confidence
As written, the strategy niftily bridges Northern Ireland’s legislative challenges.
But delivery may still be dogged by low public confidence.
Ulster University’s AI Pulse survey, which the document cites, found that only 10 per cent of 2,000 respondents agree that government or public organisations in Northern Ireland will manage AI appropriately.
The appetite for governance is clearly there, but the confidence that it will be delivered is not.
This is the hardest problem for sub-national governments to solve.
When a government cannot tell its citizens or its businesses which rules will apply, it is not well placed to ask them to trust it.
Building governance structures that will survive no matter the legislative outcome is the right approach for Northern Ireland, and other subnational governments should take lessons here.
But it is a fundamentally a placeholder for the concrete certainty that public trust requires.
Unlike Scotland’s strategy, Northern Ireland’s document is not yet finished.
The government have called it “the first step in an important journey.”
Whether it can deliver on this journey and bring all its stakeholders with it remains to be seen.
The draft strategy and consultation questionnaire are available through the Executive Office – it is worth reading and responding to before the consultation closes on October 7.
The article was originally published on interweave.gov’s Substack here, and edited.
———————————————-
Currently a Master of Public Policy candidate at the Hertie School, the author is a Tech Policy & Digital Governance writer at interweave.gov, a digital platform and newsletter focused on fostering global public sector innovation, particularly in GovTech across Europe and Asia.
Leave a Reply