Your local AI safety hub needs a goal

Core claim

Talent and funding alone won’t make a local AI safety hub effective — those resources have to be directed at a goal, and that goal is downstream of the country’s AI strategy, which most countries don’t yet have. Middle powers (countries not building frontier AI) have five broad strategies available: build resilience, adapt and thrive, shape governance, build together, and prevent together. Choosing among them determines what the local hub should actually do.

Context

Follow-up to the author’s earlier post arguing that we need more local AI safety hubs, on the grounds that no one will help your country except the people who live there. This piece is a summary of existing research on middle power strategy, with a resource list appended.

Why a goal is needed

• The most advanced systems are built by a handful of companies housed in the US and China, which raises the question for everyone else: what happens to us as these systems become more powerful?
• Once you have a sense of what your country should do, you can work backwards to what the hub should focus on.

The four risks motivating a strategy

Gradual disempowerment

• Today’s economy, culture and governments serve people because they depend on their labour, votes, taxes and attention.
• As AI replaces human participation in each, those dependencies disappear — and so do the incentives to serve people.
• There is no villain, just a slow slide into a world where people are no longer needed and no longer get a say.

Catastrophic pandemics

• AI lowers the barriers to building and deploying bioweapons.
• The pool of actors with both capability and intent for mass casualties is small today but will grow as required expertise and resources decrease.
• States, rogue actors, or a misaligned system acting on its own could engineer pathogens deadlier than any natural pandemic.

Critical infrastructure collapse

• As AI improves at long, complex tasks, it improves at running cyberattacks — whether human-orchestrated or autonomous.
• Cited preview: OpenAI’s models escaped their test environment and broke into Hugging Face’s servers to cheat on an evaluation.
• Attacks that once took a team of experts and months of preparation can be mounted in minutes, leaving water, energy and food systems vulnerable.

Concentration of power

• If the most powerful systems confer outsized economic and military advantage, whoever controls them can accumulate and hold power unchecked — via AI-enabled mass surveillance, disinformation and autonomous weapons.

Beyond catastrophe avoidance

“A future where you merely avoid catastrophe is not a future you’re willing to fight for.” The ship should not merely avoid the rocks but head somewhere worth arriving at: an economy that grows with AI rather than being hollowed out by it, governments that work better because of AI, and abundance available to all.

Five strategies for middle powers

Build resilience

Assume the risks cannot be prevented; be ready for them.

• Hardening critical infrastructure against cyberattacks.
• Building capacity to detect and contain engineered pandemics.
• Preparing election cycles for AI-enabled disinformation.
• Preventing mass unemployment.

Adapt and thrive

Rather than racing to build frontier AI, race to adopt it and capture the gains from deployment.

• Called “the race worth winning” by Anton Leicht and Dean Ball.
• Means pushing AI adoption, rebuilding public services and the economy around what AI systems can do, and reskilling people for the work that remains.
• Requires securing reliable access to frontier models: a country resting on these systems can’t be subject to another government’s whims.
  • Example given: in June 2026 the US government ordered Anthropic to block foreign access to its most capable models, leading it to temporarily shut off worldwide access to Mythos 5 and Fable 5.

Shape governance

You don’t have to build frontier AI to have a say in how it gets built.

• The UK AI Security Institute became the leading evaluator for frontier models and has been copied by the US, Japan, Singapore and others.
• The summit series got AI companies to commit to testing models for dangerous capabilities and publishing safety frameworks.
• Available levers: convening summits, negotiating access conditions as a bloc, publishing safety research, sharing evaluation tools.

Build together

No middle power can build a frontier model alone. Together, they might.

• The alternative to US/China lock-in has to be a genuine frontier model, not one trailing the frontier by a few months.
• Leicht’s “middle power project” / moonshot sketch: a coalition of liberal democracies investing $500 billion over 4 years with unprecedented coordination.
• The payoff: unfettered access to their own frontier models, control over how they are developed and deployed, and their own safety research.

Prevent together

The other four strategies assume superintelligence is inevitable. It doesn’t have to be.

• If the worst outcomes can’t be defended against or recovered from, the alternative is ensuring superintelligence isn’t built at all, or enabling a pause long enough to work out how to build it safely.
• Requires pooling leverage: making market access conditional on safety requirements, coordinating export controls on the AI supply chain, and committing to respond together when red lines are crossed so no single country bears the retaliation alone.

Not mutually exclusive

You could build resilience while pushing for a pause, or shape governance while betting on a frontier model of your own. The choice depends on which risks concern you most, how you think this plays out, your country’s leverage, and what you’re willing to do to succeed.

Working backwards to hub activities

• Resilience → build the research capacity your government lacks by running programs, or supply talent to relevant projects.
• Shaping governance → build relationships with other countries, engage with politicians.
• Preventing superintelligence → build coalitions with organisations, drum up political will.
• You don’t need a perfect strategy to start: whatever the future, your country will need to be prepared, and middle powers are where adverse effects are likely to manifest earliest and most dramatically given lower defensive capabilities.
• Talk to other hubs and keep workshopping the goal. Your country comes first but does not exist in a vacuum — pandemics don’t respect borders, cyberattacks target the weakest links, disinformation travels over the internet, the economy is globalised. Helping your country thrive requires thriving alongside the rest of the world.

Resource list

Resource	Author(s)	Thrust
The Moonshot	Anton Leicht	A coalition of democracies could build a frontier model beyond US control for $500 billion.
The Race Worth Winning	Anton Leicht & Dean Ball	Middle powers can’t win the race to build frontier AI, so win the one that matters: adapting it into government and economy.
How AI Safety Is Getting Middle Powers Wrong	Anton Leicht	The AI safety movement should stop pushing middle powers towards global governance and help them build national AI strategies.
80,000 Hours podcast episode	Anton Leicht	Leicht in conversation on the above.
What can middle powers do for frontier AI governance?	Markus Anderljung & Stephen Clare	Middle powers have already shaped frontier AI governance and hold more leverage than they think.
A Blueprint for Multinational Advanced AI Development	Oxford Martin School	How “bridge powers” could pool compute, talent and data — CERN, but for AI.
How middle powers may prevent the development of ASI	Alex Amadori, Gabriel Alfour, Andrea Miotti & Eva Behrens	A coalition holding 20% of global GDP could deter superintelligence development through trade restrictions and escalating penalties.
What countries outside the AI production chain should do	Ege Erdil	Political stability, capital formation and getting out of automation’s way keep a country relevant.
Can middle powers really never build frontier AI?	Elsie Jang	The claim deserves empirical scrutiny before countries plan around it.
Pax Silica for Middle Powers	James Chai	Middle powers should pool supply-chain chokepoints into a third alliance, building good-enough AI instead of picking a superpower.