AI Governance and Community Sector Voice in Australia
When global AI labs propose pauses and policy frameworks for AI governance, civil society is often named as part of the conversation, but the community sector closest to AI’s impacts is rarely in the room. This article examines who gets a voice in AI governance, from Anthropic’s global pause proposal to the everyday reality of Australian for-purpose organisations, and why the community sector’s perspective matters at every level of the conversation.
Observations
A Pause for Some Is Not a Pause for All
On 4 June 2026, Anthropic asked the world’s leading AI labs to consider a coordinated pause on building the most advanced AI systems. The reasoning: AI is getting close to designing its own successors, and the company wants the option to slow down so “societal structures and alignment research” can keep up.
“Societal structures” is doing a lot of work in that sentence. Whose structures? Built by whom? On what timeline, with whose input?
Anthropic says it plans to bring together policymakers, researchers, other AI companies, and civil society organisations to work through these questions. Civil society is named. That is worth noting. It is also worth asking: civil society at what altitude?
Photo: Melissa Askew / Unsplash
AI Governance and Community Sector Voice in Australia
The Pace of Change
The pace, in one number.
Anthropic’s case for urgency rests on data they have published themselves.
80%
Share of code merged into Anthropic’s production codebase in May 2026 that was written by Claude. Before February 2025, that figure was in the low single digits.
Source: Shumaker, Loop & Kendrick, 2026
Other research backs this up: the length of tasks AI models can do on their own, unsupervised, has been doubling roughly every four months. Anthropic’s own view is that if AI reaches the point where it can improve itself without human oversight, that could mean massive societal disruption.
If a company is telling you the pace of change might soon outrun society’s ability to adapt, the next question is simple: which parts of society does it mean?
References
Anthropic (2026). When AI builds itself. anthropic.com/institute/recursive-self-improvement
Shumaker, Loop & Kendrick LLP (2026). Client Alert: Anthropic’s Call for a Global AI Pause. shumaker.com
AI Governance and Community Sector Voice in Australia
Who Is in the Room
Civil society, at what altitude?
Anthropic’s plan is to bring together frontier AI developers, policymakers, researchers, and civil society organisations to design a way of pausing development if needed. Read that list again. Every group on it operates at the same level: national governments, big research institutions, and the kind of civil society organisations built to sit in rooms with labs and governments, speaking their language.
That is not nothing. These organisations do important work.
But “civil society” at that level is not the same as the organisations that will actually be sitting across the table when “societal structures” are asked to absorb whatever comes next. The community legal centre. The family violence service. The food relief program running on a part-time coordinator and a volunteer roster.
Photo: Stefan Lehner / Unsplash
We have seen this pattern before, closer to home. When Anthropic visited Canberra in March 2026 to meet with Australia’s Treasurer, the same gap showed up at the national level: the conversation covered infrastructure, investment, and regulation, but not the organisations sitting closest to the communities those decisions will affect.
When AI companies come to Canberra: whose voice is in the room? →
The UBI Test Case
The UBI test case.
Want to see what happens when AI’s biggest builders do try to address “what about the people affected”? There is already a case study. It is universal basic income.
Prominent AI leaders, including Elon Musk and Sam Altman, have argued UBI is needed to deal with the economic disruption AI will cause. They present it as a way to make sure AI’s benefits get shared, not just kept by a few. On the surface, that sounds like exactly the kind of inclusion this article is asking for. Someone is thinking about the people affected.
Look closer, and a researcher writing in Frontiers in Artificial Intelligence makes a more uncomfortable point. The same people pushing UBI are also the ones who stand to gain most from AI’s continued growth. Promoting UBI can pre-empt backlash and build public acceptance of AI’s expanding role, without touching the concentration of power that is driving the disruption in the first place.
A study funded by Sam Altman, running 2020 to 2023, gave $1,000 a month to 1,000 low-income people. It helped with housing and groceries. It did not meaningfully improve employment, education, or health. The conclusion: UBI alone will not fix the deeper problems AI-driven disruption will cause.
Source: OpenResearch (2024), cited in Bélisle-Pipon, 2025
Whose Framing Wins
Who gets to decide what “inclusion” looks like.
The researcher draws on sociologist Pierre Bourdieu’s idea of symbolic violence: a quiet form of domination that works by making one way of seeing the world feel natural and inevitable, often without the people affected even noticing it is happening. Applied here: when AI elites offer UBI as the answer, the framing does some quiet work. It says the disruption is coming, here is the cushion, and the conversation about how AI develops and who controls it can carry on elsewhere.
Tying basic income to fears about AI makes the case “far more vulnerable than it needs to be.”
A better approach treats basic income as what it actually is: an anti-poverty tool, judged on its own merits, not a side effect of someone else’s AI roadmap.
The point is not that UBI is a bad idea. It is that even a genuinely good idea can be offered in a way that keeps the people it is meant to help on the receiving end of a decision, rather than in the room where it is made.
Global AI governance conversations, from Anthropic’s proposed pause on frontier AI development to discussions with Australian policymakers, often name civil society as a participant, but rarely include the community sector organisations closest to the people AI will affect. This article traces that pattern across global AI policy, the politics of universal basic income, and the everyday governance reality of Australian for-purpose organisations, and argues that AI governance for the community sector has to be built locally rather than waiting for a seat at the global table.
Photo: Rich Smith / Unsplash
From the Other Direction
What it looks like from the other direction.
For contrast, here is what it looks like when the people closest to the issue do the proposing.
In a submission to Australia’s Select Committee on Adopting Artificial Intelligence, the charity Basic Income Australia argued that AI-driven job losses risk increasing fear, anger, and social tension. They proposed a federally funded pilot of a basic income, designed with Australian universities, tailored to Australia’s own welfare and employment systems.
3.2 million
Australians, including 17 per cent of all children, currently living below the poverty line because welfare payments are deliberately kept low to push people into work. The submission argues AI-driven job losses will make this worse.
Notice the difference in scale. This submission does not ask the world’s AI labs to coordinate anything. It asks for a modest, evaluated pilot, grounded in Australia’s own systems, built with Australian universities, aimed at a problem the organisation has spent years understanding.
It is not a counter-proposal to Anthropic’s. It is not trying to be. But it is a reminder that addressing AI’s impact on people does not have to mean a trillion-dollar company convening a global summit. Sometimes it looks like a small charity making a careful, specific case to a parliamentary committee.
Same Gap, Different Scale
The same pattern, at every level.
Three conversations. A global pause on frontier AI. A meeting between a frontier AI company and Australia’s Treasurer. A staff member at a for-purpose organisation deciding whether to trust an AI-generated output.
We have made versions of this point before. When AI companies come to Canberra, the organisations serving Australia’s most vulnerable communities are not in the room, are not consulted, and are not resourced to keep pace with decisions made on their behalf. And inside organisations, the moment AI governance most needs to work is not in a policy document. It is when someone under deadline pressure accepts a fluent AI output without checking it, a moment no governance framework currently reaches.
The pattern repeats at every scale. The people closest to the consequences are furthest from the decision. And the bigger the stage, the more “inclusion” can become a gesture that keeps the existing arrangement in place rather than actually sharing power. This is the gap that AI governance in the community sector has to close, because nobody else is going to close it from above.
Only 14% of Australian not-for-profits have an internal AI governance framework, despite 76% using AI tools. The gap between global conversations about AI’s future and the everyday reality of organisations serving Australian communities is not closing. It is the same gap, just at a different scale.
The AI backlash has arrived in Australian governance discourse →
In Closing
Where the work actually starts.
None of this means for-purpose organisations should wait for an invitation to the global conversation. It will not come, not at the level where the real decisions get made, and waiting for it means starting from zero once the consequences land.
What it does mean is that the work starts locally, now. Building the language, the literacy, and the evidence base so an organisation already understands its own AI use, its own risks, and its own community’s needs, no matter what gets decided in Geneva, Washington, or San Francisco. This is what AI governance looks like when it starts from the community sector rather than waiting to be handed down to it. It connects directly to community impact areas like economic displacement and work and community cohesion.
The conversations shaping AI’s future happen at the top. The consequences land at the bottom. The gap between those two places is where governance fails. That gap will not close from the top down. It gets built from the ground up, one organisation at a time.