Deepfake technology poses an existential threat to trust

What's this about?

People disagree about whether deepfake tech could destroy trust in what people see and hear. Deepfakes use computers to make fake videos, photos, or voices seem real.

What supporters say

• Fake videos can make people doubt real clips, even when the clips show true events.
• A public figure may call a real, shameful clip fake to avoid blame.
• Scammers can copy a loved one’s voice and use it to trick people for money.
• Groups can spread fakes at key times and use old public fights to shape views.

What critics say

• We do not yet have proof that deepfakes will cause all trust to break down.
• One fake clip may cause doubt, but it may not make most people believe a false event.
• Deepfakes work best with planned sharing and deep fights that already exist.
• Tools that spot fakes can help, but no one tool can catch every new kind of fake.

The bottom line

Deepfakes can badly harm trust and make scams, lies, and blame-dodging easier. But we do not yet know if they will cause a huge, lasting loss of trust in all of society.

Deepfake technology can seriously damage confidence in what people see and hear, making fraud, political manipulation and denials of real wrongdoing easier. But the evidence does not yet show that it will cause an irreversible, society-wide collapse of trust.

The case for

The clearest danger is that deepfakes can make audiovisual evidence less reliable. Synthetic political videos may not need to convince most people that a false event happened to do harm. Experiments suggest they can also create uncertainty about whether any video is authentic, weakening confidence in political communication and recorded evidence (see Figure 1). 1

That uncertainty can work in another way: through what researchers call the “liar’s dividend.” Once the public knows that realistic fakes are possible, politicians and other public figures may find it easier to dismiss genuine, embarrassing recordings as fabricated. This can make it harder for journalists, voters and institutions to hold powerful people accountable. 2

Voice cloning presents a more immediate threat to everyday trust. The Federal Trade Commission has warned that scammers can use cloned voices to impersonate relatives, company executives and public figures. Such tactics can turn a familiar voice—a signal people often trust—into a tool for fraud. 3

Deepfakes can also become more potent when they are used as part of wider influence campaigns. Their effect may depend less on a single convincing clip than on coordinated sharing, carefully chosen timing and political divisions that already exist. Official and institutional assessments identify synthetic media as one tool available to actors seeking to shape public opinion. 5

Technology alone is unlikely to solve the problem. Detection systems can struggle when they encounter unfamiliar types of deepfakes, altered files, compressed video or adversaries who adapt their methods. That makes reliance on one detector a weak defense in a fast-changing media environment (see Figure 2). 4

The case against

Still, describing deepfakes as an “existential threat” goes beyond what the current evidence can prove. The research supports serious risks to credibility, fraud and accountability, but it does not demonstrate that deepfakes have caused a permanent or society-wide breakdown in interpersonal, institutional or democratic trust. 6

Deepfakes are also only one part of a much larger misinformation problem. Misleading captions, selective editing, false claims and coordinated online campaigns can all distort public understanding without using synthetic media. That means it is difficult to conclude that deepfakes alone are uniquely capable of causing systemic institutional failure. 7

The effects are not the same for every audience. People’s existing trust in government and their political views shape how they respond to synthetic media. Deepfakes may therefore amplify distrust that is already present, rather than produce one universal collapse in confidence.

There are also practical ways to limit the damage, although none is complete. Verification by journalists and other trusted intermediaries, corroborating evidence, institutional preparation and provenance tools can help preserve confidence. C2PA-style credentials, for example, can record a file’s origin and editing history, though they cannot verify media with no credentials or prove that the event shown is true. 8

Warnings about deepfakes also carry a risk: they may help people resist fabricated content while making them overly suspicious of real material. Human ability to spot fakes is imperfect and varies with the quality of the media and the surrounding context. But imperfect detection does not mean trust is impossible.

The bottom line

The evidence strongly indicates that deepfakes pose serious and potentially growing threats to trust. They can spread uncertainty, enable scams, help public figures deny authentic evidence and strengthen broader influence operations.

But the record does not establish that the technology is an existential threat to trust. It lacks long-term, population-wide evidence showing that deepfakes have already produced lasting collapse in social trust or democratic institutions. Much of the available work identifies risks and mechanisms rather than proving durable outcomes at scale.

The most defensible response is not complacency, nor faith in a single technical fix. It is a layered system of safeguards: detection, provenance records, independent corroboration, trusted intermediaries and institutions prepared to respond quickly. These measures cannot eliminate the danger, but they can reduce the chance that deepfakes turn existing distrust into a wider crisis.