Bolsonaro Ally Blames ChatGPT for False "President-Elect" Claim in Deleted Letter

Imagine drafting a public letter with help from an AI assistant, only to have it insert language that contradicts reality, and then blaming the software when the backlash hits. That is essentially what happened in Brazil this week, and it offers a preview of a problem we are going to see more of as AI tools become routine parts of political communication.

On Wednesday, Brazilian Senator Flavio Bolsonaro posted a letter on social media describing himself as the country's rightful "president-elect." The claim was startling on its face. Bolsonaro had lost Sunday's runoff election to leftist President Luiz Inacio Lula da Silva by more than 4 million votes. He alleged, without evidence, that the vote had been marred by fraud, a claim that echoes the unsubstantiated fraud narratives that have followed contested elections elsewhere in recent years.

About an hour after posting the letter, Bolsonaro deleted it. He quietly replaced it with a revised version that swapped "president-elect" for "leader of Brazil," a softer and less legally fraught term. Brazilian media and political observers immediately began asking why the wording had changed so fast. Was it a walk-back? A legal correction? Something else entirely?

The answer, according to one of Bolsonaro's closest allies, was technological rather than political. Paulo Figueiredo, a right-wing influencer and podcast host who acts as an unofficial spokesperson for the Bolsonaro family in the United States, said the letter had passed through multiple drafts with contributions from several people. He said Bolsonaro's younger brother, former Congressman Eduardo Bolsonaro, turned to ChatGPT to help finalize the text.

How the Mistake Happened

Figueiredo's explanation reads like a cautionary tale about how generative AI tools handle version control. He said the chatbot pulled in an outdated version of Brazil's coat of arms and, more consequentially, resurrected an earlier draft of the letter, one in which Flavio had presented himself as the true "president-elect."

"When generating the document, AI ended up including an earlier version of the text, in which Flavio presented himself as the true 'president-elect,'" Figueiredo wrote on X. He added that the mistake was only caught after he had already distributed the letter to the press, by which point the damage was done.

Think of it like asking an assistant to update a resume, only to have them accidentally staple in a page from last year's draft, one that includes a claim you never meant to make public. The tool did not invent the fraudulent claim out of nothing. It appears to have resurfaced language that already existed somewhere in the drafting process, then presented it as the final version. That distinction matters. It is the difference between an AI system fabricating a lie and an AI system failing to track which draft was supposed to be the real one.

Figueiredo said that once the error was spotted, Bolsonaro settled on the term "leader" for legal and moral reasons, avoiding the more legally charged "president-elect" language. He described the reaction among those involved as finding the mix-up "hilarious" and called the correction "nothing major." OpenAI did not immediately respond to a request for comment, nor did Bolsonaro's press officers.

Whatever the technical explanation, Figueiredo was careful to separate the wording error from the substance of Bolsonaro's position. "In the end, the essential is expressed precisely by Flavio: these elections were illegitimate, he does not recognize the result, and he will continue fighting to restore democracy in Brazil," he said. In other words, the AI mishap did not change the underlying claim. It just changed how bluntly that claim was phrased.

The timing adds another layer. Bolsonaro arrived in Miami on Wednesday after leaving Brazil, where he had refused to comply with a Supreme Court order to turn over his passports as part of a money laundering investigation. He denies wrongdoing in that case. His declaration of himself as Brazil's "leader" has so far failed to win support from key allies, according to earlier Reuters reporting, leaving him increasingly isolated within his own political camp. Lula's victory, meanwhile, has already been recognized by Brazil's federal electoral court, international observers, global leaders, and even Bolsonaro's own party.

Why It Matters

This episode is small in scale but big in implication. We are entering a period where political statements, letters, and speeches are routinely drafted with AI assistance, and where the tools doing that drafting can quietly reintroduce outdated or discarded language without anyone noticing until it is public. A coat of arms is a minor error. A false claim to be a nation's "president-elect," published and shared with journalists before anyone caught it, is not minor at all.

There is also a credibility problem baked into this kind of explanation. Blaming a chatbot for inserting a politically explosive phrase is convenient, whether or not it is entirely true, because it shifts responsibility away from the humans who chose to publish the letter without a careful final read. Readers and voters are left to take the explanation on faith, since neither OpenAI nor Bolsonaro's office has confirmed or denied the specifics.

The deeper lesson here is not really about ChatGPT. It is about the gap between how fast AI tools can generate polished, convincing text and how slowly human oversight tends to catch up. When that gap shows up in a grocery list, it is a nuisance. When it shows up in a disputed election's aftermath, it becomes part of the historical record, screenshot and shared before anyone can hit delete.