# Deepfakes, Disinformation and the Crisis of Digital Trust

Author: Norbert Biedrzycki
Published: 2023-10-15
Language: en
URL: https://norbertbiedrzycki.pl/en/deepfakes-disinformation-and-the-crisis-of-digital-trust/
Summary: On 22 May an image appeared on social media showing an explosion near the Pentagon. It was shared by accounts that looked official, picked up by some media outlets and, for a few minutes, US stock…

On 22 May an image appeared on social media showing an explosion near the Pentagon. It was shared by accounts that looked official, picked up by some media outlets and, for a few minutes, US stock indices dipped. The image was fake, almost certainly AI generated. The Arlington fire department quickly confirmed there had been no explosion.

Last month, days before the Slovak parliamentary election, an audio recording circulated in which a party leader appeared to discuss rigging the vote. It was synthetic. It was released during the preelection moratorium, when candidates and media are restricted in what they can say.

And in 2019, the chief executive of a British energy company transferred around 220,000 euros after a phone call with someone he believed was the head of his German parent company. The voice, including its accent, had been generated by software. At the time it was a curiosity. Today the tools that made it possible are widely available and cheap.

## The end of seeing is believing

For most of history, the cost of convincingly faking a photograph, a video or a voice was high. That cost protected a basic social assumption: if you saw it or heard it, it probably happened.

Generative AI has reduced that cost almost to zero. The consequence is not only that fake content will spread. It is that genuine content will be doubted. Researchers call this the liar's dividend. When anything can be fabricated, anyone caught on camera can claim it was fabricated.

## The corporate exposure

For businesses, deepfakes are not only a political problem. They create concrete risks.

**Fraud.** Impersonation of executives to authorize payments, change supplier bank details or extract confidential information.

**Market manipulation.** Fake statements, images or announcements that move a share price for minutes, which is long enough for someone to profit.

**Reputational attacks.** Synthetic recordings of executives saying things they never said.

**Erosion of internal trust.** If employees cannot trust a voice message from their CEO, every communication requires verification.

## What organizations can do now

Technology will help. Initiatives such as the Coalition for Content Provenance and Authenticity are developing standards to attach verifiable information about the origin of digital content. Watermarking of AI generated material is being discussed by regulators and developers. But these tools will take years to become universal, and they will never be complete.

In the meantime, the most effective defenses are procedural. Payment authorizations that never rely on a single channel. Verification codes or callback procedures for unusual requests. Training that makes employees comfortable questioning a senior voice. Crisis plans that assume a convincing fake of the CEO will appear one day.

In other words, organizations need to stop treating sensory evidence as proof and start treating it as one signal among several. That is a significant cultural shift, and it is better made deliberately than after an incident.

> *Leadership question: How can organizations preserve trust when seeing and hearing can no longer be considered reliable evidence?*
