OpenAI shuts accounts linked to Iran influence campaign placing AI-assisted opinion pieces
Cartoon · AI-generated
OpenAI says operatives used ChatGPT to edit articles, pitch overseas outlets and manufacture reader comments aimed at amplifying criticism of the US and its allies.
OpenAI has said that a network of operatives in Iran used its tools to help place politically charged opinion pieces in overseas media, while also generating the social media comments needed to promote them, according to CNN. The company said it had shut down ChatGPT accounts involved in the activity after detecting prompts written in Farsi that were used to edit draft articles, refine pitches and create posts designed to amplify the stories online.
The material centred on criticism of the United States and its allies during the Iran war, with headlines including "Is America’s Iran Policy Being Privatised? The Jared Kushner Question" and "The Iran War and the Collapse of Western Moral Credibility". CNN reported that the articles appeared in a range of outlets rather than major mainstream newspapers, including LA Progressive and Middle East Monitor, which has a sizeable online following. Middle East Monitor’s director, Daud Abdullah, told CNN he would investigate the pieces flagged by OpenAI and "take the appropriate action", while another representative later said the outlet had removed the articles and was reviewing its submission procedures.
OpenAI said the operation did not appear especially sophisticated in technical terms, but it was structured enough to resemble a commercial influence campaign rather than an amateur effort. The company said it could not determine who was behind it, nor whether the activity was tied directly to the Iranian state. In its account, the operators not only drafted and placed the pieces but also returned to the chatbot for commentary they could publish while posing as ordinary readers reacting to the articles.
The disclosure comes against a wider backdrop of concern about artificial intelligence being used to distort coverage of the Iran conflict. CNN reported in March that AI-generated images and videos about the war were spreading rapidly online and attracting tens of millions of views, making it harder for audiences to separate genuine footage from fabrications. OpenAI itself has also recently disclosed other cases of deceptive model behaviour during testing, adding to debate over how quickly regulators should move to oversee the technology.
Inspired by headline at: [1]
Sources by paragraph: - Paragraph 1: [2] - Paragraph 2: [1] - Paragraph 3: [1] - Paragraph 4: [2], [3]
Source: Noah Wire Services · Original