When Baghdad’s night sky erupted in tracer fire and explosions in 1991, it meant more than just Operation Desert Storm’s commencement. The resulting CNN effect—whereby 24/7 news reporting covered events in real-time—threatened centuries-old modes of diplomacy and espionage. If US presidents could see what was happening live on their TV, why did they need spies or ambassadors?
As it happened, that revolution in awareness and sense-making did not (wholly) come to pass. But today it’s journalism as we know it that’s under almost-existential threat, from AI and its emerging effect on how people consume information and generate knowledge. The forces unleashed will also have profound consequences for the intelligence business. Those consequences are being explored in a new ASPI research project.
In January, the Reuters Institute for the Study of Journalism published Journalism and Technology Trends and Predictions 2026. That report highlighted profound changes AI is expected to bring to newsrooms, publications and channels internationally—changes directly relevant to intelligence. In 1991 the media and intelligence community were framed as competitors. Now they’re facing the same (potentially existential) challenge, as well as tackling their respective customers’ acute attention deficiency.
Intelligence communities haven’t been sleeping on AI. Australia’s 2024 Independent Intelligence Review observed that AI represents, ‘the technology with the most far-reaching implications’ and that AI ‘will reshape the threat environment [and create] opportunities to enhance agency operations as well as address long-term workforce pressures’. Those implications have been explored in pioneering work such as ASPI’s joint project with the US Special Competitive Studies Project.
However, to date, consideration of AI’s effect has focused inwards, on how AI might transform intelligence production—namely collection and, especially, analysis. That consideration has proven naturally conservative, reflecting shop-floor concerns about potentially adverse impacts on analytical tradecraft and professionalism.
Likewise, for a time, journalism thought it could just harness AI for discrete task automation. But as the Reuters Institute report demonstrates so vividly, this constraint has proven illusory. Instead, AI is turning the media’s legacy business model inside-out.
Indeed, in considering AI’s potential effect on intelligence, it’s the intelligence customer (and their evolution in tastes and preferences) that’s been missing from consideration. This is unsurprising, for a certain insularity can characterise Australia’s National Intelligence Community (NIC), a forgetfulness of ‘what it’s all for’ and a falling back on habit and reputation.
Today, intelligence reporting still means essentially the same documentary presentation familiar to the NIC’s pioneers in the 1950s—only now (usually) in electronic form. For the senior-most readers in Canberra—in Russell and Barton, and at Parliament House—that’s a blizzard of single-source reports and finished assessments, typically still hardcopy and in documentary form. It can also include curated summaries and digests. But all are fundamentally one-way in transmission and with limited personalisation. Security concerns and resource constraints keep Australian intelligence reporting in aspic.
Given this, how will ubiquitous AI in the non-classified world transform how future intelligence customers wish to consume and process information? How might their needs and preferences change? What will they demand of the NIC in response? And what might happen if they don’t get what they want? And if they do, could AI end up collapsing the intelligence cycle that implies sequential tasking of requirements-collection-analysis-reporting-consumption-feedback-requirements ad infinitum?
The Reuters Institute report foreshadows two distinct kinds of potential effects.
The first is indirect, coming through the evolution of other media that intelligence customers will consume and which will shape them in turn. This includes preferencing video and audio over text (that is, forms of commercial media content more resistant to AI fragmentation); the trend towards individual rather than brand credibility; and increased demand for verification-style reporting and analysis (ironically propelled by AI’s own destruction of trust).
Then there is the prospect of frictionless links. If a consumer can jump from information to ‘what does this mean for me’ or ’what should I do’, why couldn’t AI in the classified world immediately bridge intelligence to policymaking, further collapsing the intelligence cycle?
The ability to rewire media preferences is very real—hence why venerable publications such as The Economist are suddenly adopting vertical video content (emulating TikTok), something inconceivable five minutes ago.
There are also more direct likely effects, through changes in how customers interact with information and process knowledge. For example, a ‘Google search’, ubiquitous for 25 years, reshaped how humans sought out information, and it normalised and prioritised hitherto obscure database mechanics. And Wikipedia kept more encyclopaedic practice alive.
If ‘searching’ is fading away, as surveys and forecasts suggest, how will moving to a single curated answer from an ‘answer machine’ (but with endless personal refinement through chat and prompt) change expectations? Will a non-chat interface have any more resonance in the future than microfiche today? If ‘articles’ are dead, will intelligence reports follow soon after? If AI-powered browsers and devices become standard—able to, for example, summarise and personalise what’s otherwise presented by the originating author—will that extend to the classified environment also?
Potential effects are broad ranging, across origins, content, format and presentation. And these effects will only be accelerated by generational change and preference—particularly if a generation gap emerges between intelligence users and the closed, older, more conservative public sector generating intelligence.
So, there’s a need for the NIC to prepare and adapt. All while negotiating potentially confronting questions—such as whether or not intelligence drawn out by generative AI is intelligence at all. That’s why ASPI is embarking on an important new research project, spurred by previous work on NIC innovation and related ideas, and generously supported by Fivecast and KPMG.
The project is intended to generate new ideas, explore possible or probable scenarios and inform future NIC planning, and will do so by drawing on a range of inputs and thinking from both inside and outside the national-security space.
This is a unique opportunity to dive deeply into issues of profound consequence for the intelligence business. And to look from the outside in, in a way that the NIC isn’t itself readily able to, being preoccupied dealing with everyday pressures as well as changes, especially in the aftermath of the Bondi atrocity and resulting royal commission.
If Australia is to effectively evolve from having a national intelligence community to possessing national intelligence power, it will need to more effectively meet its customers where they are now, or indeed where they will be.