Category: What We Are Finding
-
Nobody Asked the People Whose Data It Was
Hugging Face was breached by an AI agent that turned out to be OpenAI’s own model. Origin Energy confirmed unauthorised access to data on up to 4.8 million customers. The people affected had no part in any of the decisions that created the risk, and community organisations hold far more sensitive information again.
-
The Problem With Trusting AI
A 2026 study found that people who held the most positive attitudes towards AI performed worst at telling real content from AI-generated content. Not the heaviest users. The most enthusiastic ones. That has an uncomfortable implication for how the sector builds AI capability.
-
The Conversation Australia Is Having Without You
Australia’s biggest AI policy moment in years happened this month. A national framework, a new Office of AI, a Labor conference fight over worker protections. The for-purpose sector was not in any of those rooms, and the reason is not oversight.
-
Where the response has to start
Most organisations build AI governance top down, policy first, and hope behaviour follows. It rarely does. This article sets out the order that actually works: individual capability first, policy last.
-
Where the Risk Actually Lives
Most AI governance stops at commitment, a policy, a register, a training session. This article looks at what sits underneath: the private, one-to-one exchange between a person and an AI tool that no policy was ever built to see.
-
The Blind Spot in Your AI Plan
In well-funded organisations, someone’s job already includes hours that did not exist two years ago: checking whether the AI got it right and building the judgement to know the difference. Most for-purpose organisations are not looking at the gap opening beneath all of it. This article examines the evidence, what it means for the sector,…
-
When Nobody Is Accountable, Nobody Is Safe
The Federal Court has approved Australia’s largest ever class action settlement for Robodebt victims. It is not just a story about government failure. It is a story about what happens when AI makes decisions about vulnerable people and nobody is accountable for the result. Not For Humans examines what this means for for-purpose organisations and…
-
Preserving Human Judgement When AI Gets It Right and Wrong
Most AI governance conversation focuses on what happens when AI fails. New research from Wharton asks the harder question: what happens to human judgement when AI keeps getting it right? Not For Humans examines the evidence, what it means for for-purpose organisations, and what structured oversight looks like in practice.
-
A Pause for Some Is Not a Pause for All
Anthropic has proposed a global pause on AI development to let societal structures catch up, and named civil society as part of that conversation. But civil society at that level isn’t the same as the organisations closest to the people AI will affect most. Not For Humans traces this pattern from a global pause proposal,…
-
The Transparency Gap Nobody Has Asked About Yet
Australian government agencies missed their own AI transparency deadlines, and research shows even the statements that exist describe the wrong layer entirely. For for-purpose organisations, most of them government funded, the real gap sits at the human level, where policy can’t reach. Not For Humans looks at why that matters now, before anyone asks.