The AI Exponential Is Coming to Your Council Chamber

In June 2026, Anthropic CEO Dario Amodei published Policy on the AI Exponential, an essay arguing that artificial intelligence is improving so quickly that our policy institutions — built for a slower world — are at risk of being left behind. He opens with an image from The Lord of the Rings: the Ents, ancient tree-beings so slow that a simple greeting takes a day, being asked to respond to an army that moves in hours. AI, he argues, is the army. Government is Treebeard.

Amodei writes primarily for Washington. But if you sit in an Australian council chamber, an executive leadership team, or a community services directorate, you should not file this essay under "federal issues, someone else's problem." Much of what he describes will land first, and hardest, at the local level — in our labour markets, our planning systems, our libraries and our communities. Here is what the essay says, and what I think it means for the 537 local governments that deliver services to every Australian, every day.

The argument in brief

Amodei's core claim is that AI capability is compounding exponentially. In four years, models have gone from barely writing a line of code to writing most of the code inside frontier AI companies, with similar leaps in law, finance, biology and translation. If the trend continues even a year or two longer, he expects something like "a country of geniuses in a datacenter" — AI systems that can perform most cognitive work at or beyond expert human level.

On that basis, he argues the era of light-touch AI policy is over, and sets out five priorities: binding safety regulation of frontier models, modelled on aviation; serious preparation for AI-driven job displacement; reform of slow regulatory systems so AI's benefits (especially in medicine) aren't bottlenecked; protection of civil liberties against AI-powered surveillance and autonomous systems; and a coalition of democracies to secure the technology's supply chain and set its rules.

Frontier model regulation and chip export controls are matters for Canberra and Washington, not for council chambers. But three of Amodei's five themes run straight through local government. A fourth — the question of who watches the watchers — should be on every council's risk register already.

Jobs: councils are on the front line twice over

The essay's most confronting section concerns work. Amodei argues that AI may be a far more general substitute for human cognitive labour than any previous technology, and may transform the economy faster than workers and institutions can adapt. He is explicit that enduring job displacement would be dangerous and should be minimised — but equally explicit that it may happen anyway, and that pretending otherwise helps no one.

Local government sits on the front line of this twice over.

First, as employers. Councils are among the largest employers of administrative and knowledge workers in regional Australia — planners, rates officers, records staff, customer service teams, compliance officers. Much of this work is exactly the kind of cognitive labour AI is learning to do. The choice councils face is the one Amodei urges on the private sector: use AI to do more with the workforce you have — faster development assessments, shorter call queues, better-maintained assets — rather than treating it purely as a cost-cutting exercise. Councils that frame AI adoption around service uplift and staff augmentation will carry their workforce and their community with them. Councils that frame it around headcount will not.

Second, as stewards of local economies. Amodei calls for governments to seriously measure AI's labour-market effects, and to fund transition supports: wage insurance, retraining, employer retention incentives, and infrastructure that matches displaced workers to new roles. In Australia, policies like these will be designed federally — but they will be delivered locally, through the libraries, neighbourhood houses, regional development organisations and employment hubs that councils run or anchor. Local government should be positioning now, individually and through ALGA and the state associations, to be the delivery layer for whatever transition support emerges — and to make sure regional and outer-suburban communities are not an afterthought in its design.

Data centres: the lightning rod for a bigger conversation

Amodei makes a sharp observation about data centres: public hostility towards them, he suggests, is often a symbol or outlet for much broader anxiety about AI — about jobs, fairness and who benefits. His view is that AI companies should absorb any electricity price increases they cause, and Anthropic has pledged to do so.

Australian councils are already living this. Data centre development applications are multiplying across Western Sydney, Melbourne's fringe and South-East Queensland, bringing genuine questions about energy demand, water use, land, noise and what the community actually gets in return. The essay's insight is that a council cannot resolve a data centre controversy through the planning assessment alone, because the planning assessment is rarely what the controversy is really about. Councils that pair these decisions with honest community engagement about AI's broader local impacts — and that negotiate hard for community benefit, local employment and infrastructure contributions — will fare better than those that treat each DA as a one-off land-use question.

Surveillance, fairness and the council's own use of AI

Amodei warns that AI-powered analysis of everyday data could reveal the intimate details of citizens' lives at a scale existing civil-liberties law never contemplated, and argues for hard limits on how governments use these capabilities. He also proposes a principle worth sitting with: anyone facing adverse government action should have access to AI at least as capable as the government's own.

These are not abstract concerns for local government. Councils operate CCTV networks, smart-city sensors, parking and compliance systems, and are rapidly acquiring AI analytics to sit on top of them. Every one of those procurements is a civil-liberties decision. And as councils begin using AI in compliance, enforcement and planning assessment, they need an answer to the fairness question: what does procedural fairness look like for the resident on the other side of an AI-assisted council decision? Getting ahead of this — through transparent AI-use policies, privacy impact assessments, human review of adverse decisions, and clear public communication — is cheap now and very expensive later.

Don't be Treebeard

There is one more lesson, and it is the essay's framing device itself. Amodei's worry about national institutions — that processes designed for a slower era will simply be overwhelmed by the pace of change — applies with full force to local government's own systems. Development assessment timeframes, procurement cycles, customer service models and workforce plans were all built for a world that changed slowly. That world is ending.

The window Amodei describes is real at the local level too. Community awareness of AI is high, concern is genuine, and the appetite for institutions that respond honestly and competently has rarely been greater. Councils are the level of government Australians trust to be closest to them. On AI, that proximity is an asset — but only for the councils that wake up before the forest is gone.