Artificial Intelligence platform OpenAI on Thursday said that it has identified and removed five covert influence operations based in Russia, China, Iran and Israel that were running various campaigns including anti-BJP and pro-Congress campaigns using AI. The US-based company described how these groups are using the company’s AI tools for various disruptive activities like generating social media comments, articles and images in several languages, creating fake persons, debugging codes and translating text.
As per OpenAI, these operations targeted various issues like the Russia-Ukraine war, the Israel-Hamas war, dissidence in China, politics in India, Europe and the US. The operations used various social media platforms like X, Facebook, Telegram, Medium, Blogpost etc to post the content in an attempt to sway public opinion in specific directions. However, OpenAI noted that these influence operations were not successful as they didn’t manage to engage a substantial audience.
As per the OpenAI statement, the influence operations that targeted the Indian General Elections and tried to spread anti-BJP and pro-Congress propaganda were run by an Israeli political campaign management firm named STOIC. OpenAI called this operation Zero Zeno. “The people behind Zero Zeno used our models to generate articles and comments that were then posted across multiple platforms, notably Instagram, Facebook, X, and websites associated with this operation,” the AI company said.
As per the statement, STOIC used AI tools to spread anti-Hamas, anti-Qatar, pro-Israel, anti-BJP, and pro-Histadrut content across the internet. This shows a conflicting direction, as BJP is seen as an anti-Hamas and pro-Israel political party.
The Zero Zeno operation has been targeting audiences in Canada, the US and Israel with content in English and Hebrew. But in May, it started targeting audiences in India with English content. OpenAI says that the operation used its models to generate web articles and social media comments which were then posted across multiple platforms, notably Instagram, Facebook, and X.
Some of the accounts posting the messages were originally identified by Digital Forensic Research Lab, an open-source research venture of the Atlantic Council. As per the report, the operation first targeted the rise of radical Islam in Canada, as shown in screenshots of several tweets in the report. It also targeted American Universities accusing them of promoting pro-Palestine protests, antisemitism and terrorism. After focusing on Qatar and Israel, the group then started using AI to generate content on India in May.
As per OpenAI, Zero Zeno generated content using its tools that criticised BJP and praised the Congress party.
The report also states that the operation used the same accounts to post content on different subjects like accounts that targeted radical Islam or pro-Palestine protests in Canada and the US switched to India or Israel.
OpenAI report includes some of the tweets posted as part of the campaign, masking the account’s identity. However, the tweets which are not deleted can still be traced using the text of the tweets.
One of the accounts is named Ankur Mani, with the account @ManiAnkura12515. A tweet dated 4th May, saying “No more support for BJP as long as Amith Malvia is involved” from the account is included in the report. A cursory glance on the account shows that it is posting and reposting pro-Congress content on 𝕏.
Another tweet by an apparent Indian account cited in the report is by an account named Mohini Krishnamoothi. However, the tweet talked about building applications with the MERN stack. As per OpenAI, this was posted to build a persona. This was the first tweet at the beginning of May while starting to focus on India. After that, the account posted content related to Canada, reposting a tweet by United Citizens for Canada (@canada_ucc) which is the main Canada-focused account of the network as per OpenAI. After this, the group shifted its focus to India.
Some of the comments were also posted as replies to prominent personalities, and the replies had no relation to the original post.
The impact assessment by OpenAI states that Zero Zeno’s activity appears to have attracted little if any engagement, other than from its own inauthentic accounts. It added that many of the accounts are already disabled by Meta and X, and therefore the current engagement figures may not present the complete picture. The report added that the campaign was not amplified by people outside the network.
