Porn Chatbot Tricks Argentinians Into Thinking They’re Chatting With President

Is that person you're interacting with online a real flesh-and-bones human, or are they actually just made of 1s and 0s?

Sometimes, it's easy to tell that Twitter egg that speaks like a robot is, indeed, a robot. Other times, it's trickier. Sometimes, you're really reading tweets from the official account of your country's president. Other times you're reading tweets written from a recycled pornbot pretending to be your president.

Since last year, hundreds of Argentinians have interacted with a Twitter account purporting to belong to controversial then-President Cristina Fernández de Kirchner. The account displayed her name and a picture of her, and the account's handle (@cfkresponde) and bio were similar to those of Kirchner's official Twitter account (@cfkargentina).

The account has more than 8,000 followers and several of its followers appear to have tweeted at the bot believing it to belong to Kirchner.

"It's been 30 years that I've been building democracy day by day from being a radical militant, others build personal wealth using votes," someone named Lucrecia Menor tweeted in Spanish at the unofficial Kirchner's account, @cfkresponde, weeks after Kirchner left the presidency.

"Your notions of temporality are a bit confused. Very middle class," @cfkresponde quipped back.

Lucrecia Mayer didn't back down. "Is it bad to want everyone to be equal with the same opportunities? You are the one who's confused," she tweeted.

"You shouldn't be that extremist," @cfkresponde finally said.

This seems like a realistic—if despondent and irritating—interaction. However, Lucrecia Mayer wasn't really addressing Argentina's former president, nor any of her assistants. In fact, there was no human being behind the keyboard. The unofficial Kirchner account was not only an unsanctioned fake; it was a bot powered by a repurposed porn chatbot from the 90s.

"For months and months nobody—not even one person—said: 'no, you're a bot, this is stupid,'" said Roni Bandini, a writer, journalist, and coder from Buenos Aires, who created @cfkresponde at the end of summer 2015.

Bandini programmed the bot to mine Kirchner's speeches and some of her supporters' Twitter timelines. Based on that database, the program wrote regular tweets as well as directly responding to people. Until he wrote a short anonymous article in the hacker magazine 2600 last spring, outing his project, it seemed like no one ever figured that it was powered by code.

Plenty of people doubted Kirchner herself was typing the tweets, or that it was even an official account run by some aide, but no one thought it was a bot, according to Bandini. He only spoke at length about the project, and revealed that he authored it, during a talk at the New York City conference Hackers on Planet Earth, or HOPE, this summer.

Despite fooling so many people, the code wasn't that sophisticated. Bandini explained that he simply recycled a bot that he had created in the late 1990s for a porn website designed to give people a chance to chat with supposedly-real women and men in an attempt to convince them to strip naked.

At the time, erotic webcam sites and streaming porn was still years away. And while some people did pay to chat with the apparently attractive men and women and their picture avatars, the website was a failure—a "complete disaster," as Bandini put it.

"People saw the trick, they weren't really convinced that there was a person behind the screen," Bandini told me.

So Bandini shut down the website and forgot about the bot for years. Then last year, while he was reading Kirchner's real Twitter account, which she used more often than press conferences to address the nation, it just clicked.

"I thought, 'well, my pornbot is able to tweet like that!'" Bandini joked during the talk at HOPE.

To his surprise, "things started to work really well." The same code powering a failed pornbot tricked plenty of people almost 20 years later, engaging hundreds of Twitter users in short, sometimes more or less intelligent, conversations.

The only difference, Bandini explained, is that the bot now has a "huge database." Other than that, he said he barely changed the underlying code, simply migrating it to more modern coding languages, programming it to connect to the Twitter account, and cleaning up the responses to remove some of the more vulgar vocabulary left over from the 90s pornbot. Bandini also loaded the bot on a Raspberry Pi that would automatically monitor the account and send out responses.

"The USAN [The Union of South American Nations] perhaps doesn't have the backstage of the United Nations nor the marketing of the United Nations but it's useful," Kirchner's bot tweeted in Spanish.

A user named Mariano Suarez responded: "QUESTION…goodness gracious…will you use working clothes or you'll have a designer dress you?"

"You speak about clothes because you must envy how well I bear myself," the bot answered back.

For Bandini, the experiment was a success, but also a source of concern.

Initially, his concern was visceral. What if the government found out who was behind this and didn't like it? What if they went after him somehow? This might sound like a far-fetched concern, but as he reminded during his conference talk, during the Kirchner administration, a famous prosecutor was found dead in an apparent suicide the night before a much-anticipated testimony in front of the country's Congress, where he was expected to accuse the President of covering up Iran's alleged role in a bloody bombing in 1994.

"As we say in Argentina, we know it was a suicide—we just didn't know who did it," Bandini joked.

For all he knows, however, no one in the Kirchner administration ever found out about the bot, and if they did, they didn't care. Now that Kirchner isn't president anymore, his concern is what this Twitter presidential bot teaches us about humanity in 2016.

@cfkresponde is the latest, though certainly not the last, example of a Twitter bot that pretends to be human. In our day and age, whether the online accounts we interact with are humans or not will become an increasingly relevant question, especially considering the countless online bots peddling questionable diet pills, spreading government propaganda, and even spouting racist and white supremacist proclamations on Twitter.

"Why is it that we cannot tell the difference between a person and a bot today?" Bandini asked rhetorically. "Twenty years ago this same application, the average computer user wasn't convinced about this illusion. And today, it's really easy."

"Maybe we can't tell the difference anymore because we behave today as bots," he added. "We write as bots, we use small phrases, we don't pay attention to the other part. So we cannot tell the difference anymore because a bot writes just like us."

Want more Motherboard in your life? Then sign up for our daily newsletter.