The Accelerants Making Disinformation Unstoppable

Propaganda is old. This is not.

A fair sceptic might shrug: we have always had propaganda. Gartner’s Dave Aron, co-author of World Without Truth, agrees the tactic is ancient, the book “even has an example from the Roman era.” What has changed is the machinery. Aron points to three digital accelerants and one that is not digital at all, and together they are what make this moment different, and what turn disinformation into an enterprise security problem.

1. Deepfakes

The first is the obvious one: “scary-good deepfake audio, pictures, and even video now, that’s very hard to tell from the truth.” Synthetic media has crossed the threshold where a human cannot reliably tell real from fake, which means seeing is no longer believing for any high-stakes content.

2. Agentic AI that needs no human

The second is reach plus autonomy: “the combination of global reach of social media plus agentic AI that can reach and do a digital disinformation campaign without even bothering a human.” The cost and effort of running a campaign collapses toward zero. A single operator can now field what used to take an organisation.

3. Mass-customised narratives

The third is the one Aron calls “perhaps the scariest.” A skilled disinformer can now run “a billion-person campaign, but every single person gets a slightly different message”, tuned to what is known about them. As he puts it, “that’s never been possible before.” It is the end of the shared falsehood: instead of one lie to debunk, there are a million bespoke ones. It is also why malinformation, true-but-twisted framing, scales so easily now.

4. The non-digital accelerant: your own mental model

The fourth is human. We are, Aron says, “much better at behavioural science now”, informally, nudge theory. And the most advanced attackers use it ruthlessly. The scary finding from his research: “the best disinformers, they don’t even try and give you bad facts. They try to pollute your mental model, so you make your own bad facts.” That is far more durable than any single lie, because the target does the rest of the work themselves.

Why content-based defence loses

Stack those four together and the implication is stark. If content is synthetic, infinitely customised, and machine-generated at scale, then any defence that works by inspecting content is chasing an infinite, ever-changing target. You cannot fact-check a million bespoke messages. You cannot fingerprint content that is unique to each recipient. You cannot out-read a machine that writes faster than you.

But all four accelerants share one thing they cannot hide: coordination. A billion customised messages still need to be launched together, by connected accounts, on synchronised timing, toward a shared goal, the same signature behind any coordinated campaign. The content varies infinitely. The orchestration does not.

That is the case for reading behaviour instead of content, and it is what AI Uniti is built on. Signal scores the coordination, the timing, network and account signals, behind a campaign and raises explainable alerts regardless of what the content says or whether it is synthetic. PulseCheck identifies the inauthentic accounts at scale. When the content is designed to be unbeatable, you change what you look at: not the message, the machine sending it.

When the content is engineered to be unbeatable, change what you look at. See how Signal reads the coordination behind a campaign, not the content, or book a 15-minute demo.