# The industrialisation of disinformation: a new era on the horizon

Author: Fernando Nieto Lobato
Original publication: 2023-10-04
Spanish original: https://estrategiabyaleph.substack.com/p/estrategia-1
English URL: https://elcontemplador.github.io/estrategia-english/essays/001/
Status: Published translation

This is a translation of the original Spanish essay published on 4 October 2023. Its claims, examples and forecasts retain that historical context.

English publication: 2026-09-29

“***Please remain. You furnish the pictures, and I'll furnish the war.***” Whenever disinformation comes up, those **supposed words** (which may well themselves be apocryphal) almost immediately spring to mind: **the telegram** from newspaper magnate **William Randolph Hearst** (so magnificently brought to us by Welles in *Citizen Kane*) **to his illustrator in Havana**, after the artist had asked to return to the United States because nothing was really happening in Cuba. Yet days later, just as Hearst himself had asserted, the United States entered the war in 1898.

For although **disinformation and the manipulation of information, including their political use, have been with us since time immemorial**, **technology has caused the number of cases to soar** over recent decades, and particularly in recent years. Examples include [Cambridge Analytica](https://www.bbc.com/mundo/noticias-43472797), especially its significance for Trump's 2016 campaign, and what happened around Brexit. **The point is that, as generative artificial intelligence becomes widely available, disinformation will very probably become far more pervasive still, to the point where we may soon be unable to believe anything we see or hear unless we witness it in person.**

**In his very recent book *The Coming Wave***, due to be published in Spain in a week's time, **Mustafa Suleyman**, founder of DeepMind, Google's artificial intelligence laboratory, devotes an interesting chapter to these issues and **coins a term** that particularly caught my attention: **“the industrialisation of disinformation”.**

Until now, producing highly convincing manipulated information required professional work and advanced tools that were not generally available to everyone; or, as in Hearst's case, control of a mass-circulation newspaper and very little regard for the truth. **We can now see how anyone with a simple mobile app will be able to create *deepfakes* that look entirely real to a non-specialist**, whether already today or in the months or years ahead. Following a distressing case in Spain a few days ago, in which fake nude images of minors were circulated, one expert warned about the danger of “one-button apps”, which allow anyone with almost no knowledge to create fakes that cause enormous harm to others.

In the coming years, we will very probably see even candidates for mayor in small towns, or their advisers, able to generate **fake statements by other candidates**, with barely any technical knowledge, by cloning their voices. They will be able to create **fabricated news stories** on fake versions of respected websites replicated with AI, or **even fake but entirely convincing videos showing their rivals committing illegal acts**, simply by typing a few lines of text. These technologies are already available to anyone in an early form and will become more widespread and sophisticated in the years ahead. Combined with mass, anonymous distribution through social media, this could have terrible effects, particularly in local politics, where public scrutiny is much more limited.

**This past week, for example, an [AI video inpainting tool](https://shangchenzhou.com/projects/ProPainter/?ref=aiartweekly) was presented that directly removes watermarks from any video**. Watermarking is one of the precautions used by some AI video generators to identify their output.

[![A post demonstrating ProPainter, the AI video inpainting tool discussed in the article.](https://elcontemplador.github.io/estrategia-english/assets/images/7f0b2fab-298f-4b81-bf43-703b7f431961_806x876.png)](https://x.com/heyBarsee/status/1707324714710909273?s=20)

In response, companies such as [Google are trying to develop their own **systems for mandatory marking of everything generated with artificial intelligence**](https://www.xataka.com/robotica-e-ia/google-ha-creado-primera-marca-agua-para-ia-invisible-para-ojo-humano-no-para-maquinas). But these systems would need worldwide adoption, which does not look easy. As with text generated by GPT-4, which the company itself admits cannot be reliably distinguished from something written by a human, **it will be very difficult to prevent these special safeguards from being easily defeated simply by combining several AIs and manipulating the output again**. Indeed, given how hard it may be to establish that safeguard, **some are already advocating the reverse approach: marking real photographs and videos produced by cameras**, so that they can be distinguished from synthetic output.

We therefore seem to face a very complicated future for political disinformation. AI itself has been, and continues to be, used to detect and combat *deepfakes* and disinformation campaigns, on many social networks, for example. Yet its new generative capabilities may bring an unstoppable flood of disinformation in the coming years that looks highly “authentic” to human eyes. As Suleyman points out in his book, this **will make it much easier to sow instability**:

*Translation note: the Spanish article identified the following book passages as translations from English made with GPT-4. They are translated back from that Spanish text here; they have not been checked against the original English edition.*

> *Anyone motivated to sow instability now has an easier task. Suppose that, three days before an election, the president is caught on camera using a racist slur. The campaign's press office vehemently denies it, but everyone knows what they have seen. Outrage spreads across the country. Poll ratings plummet. Swing states suddenly shift towards the opponent, who wins against all expectations. A new administration takes office. But the video is a deepfake, one so sophisticated that it evades even the best neural networks for detecting fakes.*
>
> *The threat here lies less in extreme cases than in subtle, nuanced and highly plausible scenarios that are exaggerated and distorted. It is not the president storming into a school, shouting nonsense and throwing grenades; it is the president saying resignedly that there is no choice but to introduce a set of emergency laws or bring back conscription. It is not Hollywood pyrotechnics; it is supposed CCTV footage showing a group of white police officers beating a Black man to death.*

**But the threat could become far more damaging in the near future, when AI takes another step forward.** In the book, Suleyman discusses the harm that could be caused by a radicalised terrorist seeking to recruit followers for his organisation through personalised videos.

> *Soon these videos will be fully interactive and believable. You will be speaking directly to him. He knows you and adapts to your dialect and style, playing on your history, your personal grievances, the bullying you suffered at school, your terrible, immoral, Westernised parents. This is not disinformation as carpet bombing; it is disinformation as a surgical strike.*

With all these dangers already on the horizon, we are very probably entering a **new era of political disinformation**. **Perhaps the only bleak consolation is the point made by [The Economist's editor-in-chief, Zanny Minton-Beddoes, in this interesting discussion with Yuval Noah Harari and Mustafa Suleyman](https://youtu.be/7JkPWHr7sTY?feature=shared)**: “*In the analyses we have carried out [at The Economist], **we think AI may have less impact on elections and the political system in the short term because trust in politicians is already at rock bottom***.”

*Translation note: this final quotation is translated from the Spanish article's rendering of the discussion; its precise English wording has not been verified against a transcript.*

Fernando Nieto Lobato

*Director of Digital Innovation, Institución Educativa ALEPH*