New Delhi: As part of its ongoing efforts to combat the misuse of its AI models for the purpose of spreading disinformation, OpenAI has put out a report labelled ‘AI and Covert Influence Operations: Latest Trends’. The report, among other things, points to an Israeli firm called STOIC generating content using OpenAI models to generate content about the Gaza conflict, the Histadrut trade union organisations in Israel and the elections in India. OpenAI labelled the operation as ‘Zero Zeno’ to reflect the low levels of engagement.

The Zero Zeno operation has been characterised by OpenAI as a “for-hire Israeli threat actor posting anti-Hamas, anti-Qatar, anti-BJP and pro-Histadrut content across the internet.”

Modus Operandi Of The Covert Op To Target Elections in India 

1. Influence operations like Zero Zeno used OpenAI models to generate large quantities of comments and content that were then posted across Telegram, X, Instagram and other sites.
2. Some of the campaigns used their AI models to create the appearance of engagement across social media - for example, by generating replies to their own posts to create false online engagement.
3. Initially, this operation targeted social media users in Canada, the United States and Israel with English and Hebrew content. From May onwards, the operation shifted to targeting Indian audiences with generated English content. 

How Social Media Was Manipulated By The Israeli Firm

The report notes that the Zero Zeno accounts used profile pictures generated by ‘an earlier type of artificial intelligence: generative adversarial networks.’ While these images, OpenAI says, are easily downloadable from the internet, many of the accounts used the exact same profile image. 

The Indian operation, as stated above, started in May and was disrupted by the company in less than 24 hours of its conception. This campaign saw the Zero Zeno operation generate comments that focused on criticising the ruling Bhartiya Janata Party and praising the opposition Congress party. 

Furthermore, the report says that some of the content being generated by the network was posted as a response to social media posts by prominent Western figures though the responses had no relation with the actual social media posts. 

Open AI Says ‘Category 2 Operation’

As stated above, the Zero Zeno label attached by OpenAI to this alleged Israeli operation is a reference to the minimal impact made by the influence campaign. The report notes that the content generated by this operation saw very little interaction beyond what was being done by other fake accounts being operated by the network. 

Thus, on a breakout scale, OpenAI says the influence campaign was a ‘Category 2’ operation, meaning that while it was posted on several social media platforms, there is no evidence of it being amplified by people outside the operation.