This week the gap between who benefits from AI and who pays for it nearly closed globally – but the story underneath is more complicated, and more urgent, than any single number.
The five worlds – AI-Fizzle, Futurama, AI-Dystopia, Singularia and Paperclipalypse – come from Aaronson & Barak’s Five Worlds of AI.
This week's scan of nearly 300 stories from around the world found AI news almost perfectly split between good and bad futures. In the good future – called Futurama – AI is a tool that helps people do better work, and the benefits spread widely. In the bad future – called Paperclipalypse – AI causes harm that nobody can stop, like cyber attacks, fraud, and decisions made without proper checks. Right now, those two futures are in a tug-of-war. The really big story is that in Australia and New Zealand, readers are seeing more signs of AI spreading benefits than concentrating them – but in North America, the opposite is true, with lots of stories about big companies getting richer while workers get left behind. Which future we end up in depends on choices being made right now.
Here's a number worth sitting with: the global Distribution Gap – the margin by which AI-Dystopia stories outnumber Futurama stories in this week's scan – fell to just 1 point. That's the joint-lowest reading in 12 weeks of tracking, matching the scan from 31 May. After three consecutive weeks of widening (the gap hit 11 last week), this is a genuine reversal. Whether it's a turning point or noise, we'll know in a fortnight.
The global picture for the week ending 28 June: AI-Fizzle 22%, Futurama 23%, AI-Dystopia 24%, Singularia 14%, Paperclipalypse 17%. Five worlds, spread across 281 stories, almost perfectly balanced. That balance is itself the story.
But averages hide the fault lines.
With 61 signals and a 40% weighting in our global index, this region punches well above its size – and this week it delivered the most striking reading in the scan.
The Distribution Gap for Australia and New Zealand sits at –15 this week. Negative means Futurama is outrunning AI-Dystopia. That's a strongly optimistic reading, and it holds up on the rolling four-week average too: –9. Australian and New Zealand readers are, on balance, encountering more stories about AI spreading value than concentrating it.
What's driving that? A mix of genuine adoption signals and governance activity. The DTA's new AI review committee, Western Australia's data science training programme, TUANZ's call for digital workforce investment in New Zealand, the Productivity Commission's claim that AI can cure Australia's productivity slump – these are policy signals pointing toward shared benefit. Suncorp trialling agentic AI in insurance claims, Bunnings expanding its 'Buddy' agent to trade customers, The Iconic's CTO betting on Google AI for discovery – real enterprise moves, not just press releases.
The Paperclipalypse share is high at 25%, which reflects genuine anxiety: AI slop flooding Australia's financial complaints system (Capital Brief), unsanctioned AI outpacing employer guidance (HRD Australia), APRA calling for collective defence against AI cyber threats. The Five Eyes warning – eight stories in the same group – landed heavily across ANZ sources this week. The cyber risk story isn't abstract here.
The 'jobpocalypse' theme landed in two directions. Both The Mandarin and SmartCompany ran pieces citing economist evidence that mass AI-driven job displacement hasn't arrived. That's Fizzle, not comfort – the promised upside is also absent.
The contrast with North America is stark. The Distribution Gap there is +24 this week, rolling four-week average +20. That's the highest sustained positive gap in any region across our series. AI-Dystopia commands 39% of North American signals.
The biggest stories tell you why. Oracle shed 21,000 jobs in AI-driven restructuring. One in three employers surveyed by GMAC admits replacing entry-level roles with AI. The Fortune 500 is posting record revenue per employee while hiring fewer people. Sam Altman's personal investments benefiting from OpenAI ties. Anthropic and OpenAI spending $27 million in a Manhattan congressional race to shape their own regulation. Chip makers profiting at everyone else's expense. The US government restricting access to GPT-5.6, then selectively releasing Anthropic's Mythos to 100+ approved companies and agencies.
That last cluster – six stories in the same group – is the week's most structurally significant North American signal. Government as gatekeeper to frontier AI capability. Power concentrating not just in companies, but in a specific relationship between companies and the state.
There's also a genuine Singularia thread running through North America: Claude Tag (seven stories), General Intuition's $2.3 billion bet on game-trained agents, Taktile automating high-stakes financial decisions, OpenAI's internal token output growing 56x in research since November. These aren't hypothetical. The autonomy is arriving.
Europe's 24 signals produce a Distribution Gap of –12, with Fizzle elevated at 29% – the highest of any region. That Fizzle reading reflects genuine European scepticism: UnlikelyAI losing senior staff, companies buying AI licences and not using them (Tech.eu), the UK's AI minister explicitly avoiding ChatGPT for ministerial business, AI innovation outpacing regulation.
The sovereign infrastructure theme is real and consistent: TensorX/Solstice building sovereign European AI, Sifted's piece on owning the full AI stack as essential. Europe's AI story is increasingly about defensive positioning, not frontier ambition.
Rolling four-week: Distribution Gap moves from –4 to –12. The trend is toward Futurama in Europe, driven by governance investment rather than adoption exuberance.
APAC's 68 signals carry only a 5% global weight, but the readings are instructive. AI-Dystopia dominates at 41% this week, rolling four-week 44%. The Distribution Gap is +13, rolling +21. DeepSeek doubling its workforce and pursuing AGI. Alibaba tripling T-Head's capital. China's telecoms giants building multi-domain networks. Samsung's $648 billion capex plan. SK Hynix raising trillions. This is the infrastructure arms race in plain sight.
The Futurama signals exist – Indian IT generating $10–12 billion in AI services revenue, Singapore's government piloting AI translation, IHH Healthcare and Infosys on AI-powered ERP – but they're outnumbered by stories about capital and compute concentrating in national champions.
Technology (115 signals) is the engine behind both Singularia and Futurama this week – it's genuinely neutral, capable of pulling in either direction. The Claude Tag cluster, the IBM sub-nanometre chip, DeepSeek's DSpark inference improvements: technology signals are firing across all five worlds.
Economy (78 signals) is the primary driver of AI-Dystopia – Oracle layoffs, chip maker profit concentration, enterprise cloud cost inflation, the 'AI tax'. When economy drives the story, the gains are going somewhere specific, and often not to the workers.
Trust (49 signals) owns Paperclipalypse almost entirely this week: Five Eyes cyber warnings, token theft, AI slop in financial complaints, unsanctioned AI use, Workday's bias lawsuit. Trust signals are the alarm system for uncontained harm.
Talent (44 signals) cuts across Dystopia and Fizzle: entry-level jobs replaced, AI skills commanding a 62% wage premium, training gaps dulling transformation. The talent story is a distribution story.
Geopolitics (26 signals) is concentrated in Dystopia and Paperclipalypse: US government restricting model access, Anthropic versus Alibaba over illicit Claude access, European sovereignty drives, China's chip race. Geopolitics is where AI becomes a state instrument.
Three silences matter this week. First: genuine mid-market adoption evidence. The stories that move through our scan are either frontier labs, hyperscalers, governments or large enterprises. Small and medium businesses – the engine of the Australian and New Zealand economy – appear almost only as recipients of policy recommendations, not as practitioners with real stories to tell.
Second: the environmental cost is nearly invisible in enterprise AI discourse. The UN Secretary-General called out AI firms on climate impact; it generated one story cluster. For a week in which data centre infrastructure dominated, the energy and water consequences barely featured in the business press.
Third: the customer. Retail, health and financial services AI stories are mostly told from the vendor or executive perspective. The actual experience of the person at the end of the AI workflow – the insurance claimant, the banking customer, the job applicant screened by Workday – is almost entirely absent.
Here's the plain version. The global Distribution Gap just fell from 11 to 1. That sounds like good news – and in some ways it is. But the reason it narrowed isn't that AI started sharing its benefits more broadly. It's that the Dystopia signal dropped sharply (from 35% to 24%), partly because the enterprise adoption story is becoming more real and differentiated, partly because some of last week's concentrated bad news cycled out.
The signal that should follow you into the week: in North America, the gap between who benefits from AI and who pays for it is at its widest in 12 scans. If your organisation sources leadership thinking, technology vendors or investment frameworks from North America – and most do – you're swimming in a current that is normalising AI gains flowing to shareholders and AI costs flowing to workers and communities.
The question isn't whether AI is working. It clearly is. The question is whether your governance, your talent investment and your benefit-sharing model are keeping pace with the autonomy you're deploying. Because the organisations that will look best in three years aren't the ones who moved fastest. They're the ones who moved deliberately – and can explain to their people exactly who got the upside.
The articles most responsible for bending the five worlds this week.
Intelligence chiefs from the Five Eyes alliance – Australia, the US, the UK, Canada and New Zealand – told businesses and governments to treat AI-enabled cyber attacks as an urgent threat they may not be able to stop in time. Eight separate publications ran the same warning, making it the most widely covered single story of the week. For any Australian or New Zealand organisation responsible for critical systems, this wasn't background noise – it was a direct instruction to act.
Oracle cut 21,000 jobs and pointed directly at AI-driven restructuring as the reason. That's not a cautionary tale or a prediction – it's a named company, a real number, and a pattern that multiple sources confirmed this week. When a global enterprise software giant of Oracle's scale makes this move, it signals that the AI efficiency story and the job displacement story are the same story, just told from different sides of the ledger.
The US government told OpenAI to restrict who could access its most powerful new model, then selectively approved over 100 companies and agencies to use a specific Anthropic model instead. Six stories tracked this cluster. It matters because it shows that access to the most capable AI is no longer just a commercial question – it's a geopolitical one, shaped by whoever holds power in Washington at the time.
AI-generated complaints – low-quality, formulaic text created by tools people are using to chase financial remedies – are flooding Australia's financial disputes system and slowing it down for everyone. This is a local, concrete example of AI causing institutional harm that regulators weren't ready for. It's the kind of story that rarely makes global headlines but lands hardest for Australian business and policy leaders.
Anthropic launched an AI agent that lives inside Slack, reads company conversations, remembers context across time, and acts as a persistent team member rather than a tool you query. Seven publications covered the launch. Whether this is productivity augmentation or the beginning of AI taking over how organisations think and communicate depends entirely on how leaders choose to deploy it – and that choice is being made right now, mostly without formal governance in place.
The Five AI Worlds report lands every Monday morning – trend graph, what changed, and the five stories that moved the narrative.
Subscribe free