OpenAI claims that Russian agents used its AI chatbot to brief their bosses, highlighting influence campaigns that were previously not attributed to Russia.
Jeff Chiu/AP
hide caption
Jeff Chiu/AP
OpenAI said in a blog post on Thursday that it had banned Russian and Iranian accounts that used the company’s artificial intelligence tools to facilitate its influence operations in several countries. The agents said they managed to get their false narratives picked up by the media and, in one case, provoke a response from a member of the European Parliament, according to internal reports on the operations produced using OpenAI’s chatbot, ChatGPT.
OpenAI studied the impact of different campaigns mentioned by the accounts and found that those aimed at spreading false narratives in established media outlets appeared to have reached a wider audience than campaigns focused on using fake social media accounts to spread their message.
The Russian operations targeted countries in Latin America and appeared aimed at undermining support for Ukraine and influencing local politics, OpenAI said. The accounts appeared to primarily use ChatGPT to write reports detailing their work to unspecified superiors, OpenAI said. The company said one of those reports appeared to describe the largest influence operation OpenAI has seen to date, as it led to a public response from a political figure.
During this campaign, Russian agents said they posed as a regional education authority in Peru and tricked some schools into holding activities on Ukraine featuring Stepan Bandera. Bandera is a controversial figure, admired by nationalists for leading the campaign for Ukrainian independence before and during World War II. He is vilified by others, including in Poland, because his movement sided with the Nazis and killed Poles and Jews. Peruvian media outlet Extra reported that one such event took place at a Peruvian school.
A Polish report on the school event prompted a far-right Polish member of the European Parliament to suggest barring entry to people declaring support for Bandera.
“The fake Russian was thus nourished by historical tensions between Ukrainians and Poles,” writes OpenAI in its report.
“The Russians engage in these kinds of active measures all the time,” said Darren Linvill, co-director of the Watt Family Innovation Center Media Forensics Hub at Clemson University, who studies online influence operations. “Trying to engage in narrative whitewashing, creating false stories, false personas to create or spread propaganda that aligns with Russia’s worldview and political agenda.”
Internal Russian memos created with ChatGPT also claimed that the agents were running a fake think tank in Latin America and recruiting researchers without their knowledge to interview experts and write reports. OpenAI said the effort was similar to a 2020 Russian operation posing as an independent news outlet linked to the Internet research agency of the late oligarch Yevgeny Prigozhin, which participated in efforts to influence the 2016 U.S. election.
In other cases, OpenAI said, Russian agents took credit for events for which they were not responsible. The Russians did use AI to generate fake hardware, OpenAI said, but that was only a minority of the workload.
“These Russian underlings were trying to use AI to create a shortcut and do less work in the same way that many of us use AI, but in this case it helped them get caught,” Linvill said.
While Russian agents primarily used AI to brief their superiors, Iranian agents used AI to help them deliver content in Persian and English. The campaign focused primarily on the US-Iran war and, to a lesser extent, US politics. OpenAI said the operation’s activity “appeared consistent with a commercial actor conducting a for-hire influence campaign” but was unable to identify who it was.
According to OpenAI, the Iranian accounts formed seven fake authors to pitch and polish long articles to small media outlets around the world. They have published nearly 100 articles.
The fake authors published articles in media outlets including the US-based progressive news site Daily Kos and the UK-based Middle East Monitor. According to NBC News, the Middle East Monitor published at least 21 articles written by four fake authors. The media, which did not respond to a request for comment, has since deleted the author’s pages.
The Daily Kos wrote in response to NPR’s request for comment that its community section, where one of the fake personas was posted, is separate from the staff reports. Anyone can sign up to post in the community section and posts “are not subject to review by Daily Kos staff prior to publication,” according to a disclaimer on the site. The community section contains rules for posting, including a ban on “pretending to be someone you’re not.” The character’s messages are still online.
“AI was valuable to them, more than just a shortcut,” Linvill said. “It helped them create language that… seemed authentic to the point where it was published by legitimate media outlets. And I think that’s something that an Iranian operation like this would have had a lot more difficulty with a few years ago.”
Gn world
