Jair Bolsonaro’s AI-generated appearance at his son’s presidential campaign launch exposes a new problem for election law: disclosure may tell voters that a political figure is synthetic without answering whether campaigns should be allowed to reproduce that person’s political influence at all
By The Grey Ghost
THE SIGNAL
A political figure no longer has to walk onto a stage to appear at a campaign event.
He does not necessarily have to record a video.
He may not even have to authorize the message.
Artificial intelligence can recreate the face.
The voice.
The gestures.
The speaking style.
And increasingly, the political presence.
That is what makes the controversy surrounding former Brazilian President Jair Bolsonaro more important than one synthetic campaign video.
At the July launch of his son Flávio Bolsonaro’s presidential campaign, an AI-generated version of Jair Bolsonaro appeared on screen and endorsed his son.
The video explicitly disclosed that it was synthetic.
That should have made the issue simple.
It did not.
Opposition parties challenged the video before Brazil’s electoral authorities, while Flávio Bolsonaro’s lawyers argued that the disclosure meant the material complied with election rules. Jair Bolsonaro’s lawyers said he had not authorized the recreation. Brazil’s electoral court had not resolved the central legal question when Reuters reported on the dispute. (Reuters)
That leaves election law confronting a question it has never had to answer at this scale:
If a synthetic politician is clearly identified as synthetic, can the replica still exercise the political power of the real person?
That is the signal.
Political identity is becoming reproducible.
WHAT THE MARKET IS MISSING
Most discussion about political deepfakes has focused on deception.
A fake candidate saying something the candidate never said.
A fabricated scandal.
A fake concession speech.
A manipulated photograph.
A cloned voice designed to trick voters.
Those threats remain real.
But Bolsonaro’s AI appearance exposes a different problem.
What if nobody is being fooled?
The avatar reportedly identified itself as artificial intelligence.
The audience knew the real Jair Bolsonaro was not standing on the stage.
Yet the synthetic version could still perform an extremely valuable political function:
It could endorse a candidate.
It could deliver a message.
It could evoke the personality, loyalty and emotional connection associated with the real person.
It could effectively place that political figure inside an event without requiring the person to appear.
This moves the debate beyond truth versus deception.
It moves toward presence versus replication.
Brazil already requires synthetic political material to be explicitly and prominently labeled when AI has been used. Its 2026 rules also prohibit new synthetic content featuring candidates or public figures during the critical period beginning 72 hours before voting and ending 24 hours afterward—even when that content is properly labeled. (TSE)
That tells us something important.
Regulators are beginning to recognize that disclosure does not eliminate influence.
A warning label may tell people something is artificial.
It does not make the content politically powerless.
FIRST-ORDER EFFECTS
The immediate effect will be uncertainty for campaigns.
Political campaigns suddenly have access to something far more powerful than traditional video editing.
They can potentially create synthetic appearances by politicians, celebrities, supporters and public figures without arranging a conventional appearance.
A politician could theoretically deliver customized versions of the same endorsement to different regions.
Different languages.
Different demographic groups.
Different digital platforms.
A recognizable political figure could appear at hundreds of campaign events without traveling anywhere.
And once interactive avatars become sufficiently advanced, the synthetic figure could potentially answer questions and engage directly with supporters.
That creates enormous campaign value.
It also creates enormous legal ambiguity.
Who authorized the message?
Who owns the likeness?
Who is legally considered the speaker?
Does an AI recreation constitute an appearance by the real politician?
Who is responsible if the avatar says something unlawful?
Does clearly labeling the material eliminate the violation?
Can someone who is legally restricted from communicating nevertheless exert influence through an independently created synthetic replica?
Those are no longer theoretical questions.
Brazil is dealing with them now.
SECOND-ORDER EFFECTS
The larger consequences begin when political replicas stop being one-off videos.
Today’s AI avatar is usually generated content.
Tomorrow’s could be an operating political surrogate.
Give an AI system enough speeches, interviews, videos, policy positions and personal history and it can increasingly reproduce recognizable patterns of communication.
Then add voice cloning.
Persistent memory.
Real-time conversation.
Video generation.
Multilingual translation.
Avatars.
The resulting system could potentially campaign continuously.
The real politician sleeps.
The synthetic version does not.
The real politician can address one audience at a time.
The replica could address thousands.
The real politician may speak one or two languages.
The synthetic version could speak dozens while preserving the same voice.
The real politician eventually leaves public life.
The digital version may not.
That produces a new political asset:
synthetic political presence.
And that asset could become valuable enough that campaigns begin planning for it deliberately.
Candidates could authorize official digital twins.
Political parties could maintain approved synthetic spokespersons.
Campaigns could license a prominent person’s image and voice.
Families or estates could control political replicas after someone dies.
Unauthorized groups could create competing versions.
Eventually voters may encounter political figures who exist simultaneously as human beings and as scalable synthetic entities.
At that point, the old definition of political participation begins to break down.
WINNERS
The immediate winners are campaigns capable of using synthetic media responsibly and effectively.
AI dramatically lowers the cost of producing customized political communication.
A campaign could translate messages while preserving a candidate’s voice.
Generate accessibility features.
Create clearly disclosed educational material.
Scale appearances to communities a candidate cannot physically visit.
Political communication companies, AI video providers and voice-generation platforms could become increasingly important campaign infrastructure.
There is also an opportunity for authentication technology.
As synthetic media becomes commonplace, campaigns, news organizations and voters will need stronger ways to establish which messages were authorized.
Digital signatures.
Content credentials.
Provenance systems.
Cryptographic verification.
Official campaign repositories.
The ability to prove that content is authentic may eventually become as valuable as the ability to generate it.
Election regulators could benefit as well if high-profile cases force clearer rules before synthetic political identities become widespread.
Brazil may ultimately provide other democracies with an early case study of where those boundaries should sit.
LOSERS
The first potential losers are politicians themselves.
For most of history, political figures controlled their physical presence.
AI weakens that control.
Once enough public material exists to reconstruct someone’s appearance and voice, political identity becomes technically reproducible.
That creates problems even when the synthetic material contains a disclosure.
A politician may disagree with the message.
The politician may never have approved it.
Yet voters can still see and hear a highly realistic representation delivering it.
Public figures therefore face a new type of identity problem:
Their political likeness could become an independent media asset.
Voters could lose as well.
Not necessarily because every AI political message will fool them.
The greater danger may be volume.
When synthetic production becomes cheap, political communication can become nearly unlimited.
Thousands of personalized speeches.
Synthetic endorsements.
Automated campaign conversations.
Localized messages.
Instant rebuttals.
AI-generated surrogates.
The information environment could become saturated with artificial political presence even when every piece is technically labeled.
And disclosure itself could eventually lose power.
If voters see “AI GENERATED” hundreds of times every day, the warning may become background noise.
WHAT HAPPENS NEXT
Brazil is unlikely to be the last democracy forced to confront this.
The United States is already building a patchwork of political-deepfake regulation.
As of June 2026, 31 states had enacted laws regulating deepfakes in political messaging, according to the National Conference of State Legislatures.
Most rely heavily on disclosure requirements, although some jurisdictions impose temporary prohibitions or other restrictions. (NCSL)
Brazil has gone further in several respects.
Its election authority requires prominent disclosure of synthetic content and prohibits certain newly created AI depictions of candidates and public figures during the final 72 hours before voting and the 24 hours afterward, regardless of whether they are labeled. (TSE)
That suggests the regulatory debate may evolve through several stages.
First came:
Was AI used?
Then:
Was the voter deceived?
The next questions are harder:
Did the depicted person consent?
Who authorized the political message?
Can a synthetic replica legally act as a campaign surrogate?
Can campaigns generate unlimited versions of public figures?
Should political replicas remain usable after the real person can no longer—or is no longer permitted to—participate?
And eventually:
Does an AI political identity have to be controlled by the human it represents?
Those questions will become unavoidable as synthetic media moves from prerecorded clips toward interactive agents.
The decisive moment will come when voters are no longer simply watching an AI politician.
They are talking to one.
BOTTOM LINE
The Bolsonaro controversy is not ultimately about whether one campaign video contained the correct disclaimer.
That is the immediate legal question.
The Deep Signal is much larger.
Artificial intelligence is separating political influence from physical presence.
For centuries, a politician’s image, voice and personality were tied to the politician.
Mass media loosened that relationship.
AI can sever it.
A recognizable political figure can now potentially speak without speaking.
Appear without appearing.
Campaign without traveling.
Address voters without being in the room.
And possibly exert influence without personally creating the message.
That creates a new challenge for democracies.
Election law has traditionally regulated candidates, campaigns, advertisements and broadcasters.
It may now have to regulate synthetic political identities.
Disclosure will remain important.
But the Bolsonaro case suggests disclosure may not be enough.
Because voters can fully understand that a politician is artificial and still be influenced by everything that politician represents.
The next political deepfake problem therefore may not be:
“Is this real?”
It may be:
“Who has the right to make this person politically real again?”
