OpenAI blocks Iranian attempts to manipulate US election using ChatGPT
OpenAI has identified and deactivated several ChatGPT accounts linked to an Iranian influence campaign. The campaign was primarily using ChatGPT to generate extensive articles and social media comments, on platforms like Instagram and X. The accounts were associated with Storm-2035, a clandestine Iranian operation known for creating websites that mimic political news outlets, in an attempt to sway US voters.
Content spanned various topics
The influence campaign generated content on a wide range of topics. These included the US presidential election, the Israel-Hamas war, Israel's participation in the Olympic Games, Venezuelan politics, and issues concerning Latinx communities in the US. The operation was part of a broader series of influence campaigns linked to the Iranian government by Microsoft last week.
The campaign achieved limited engagement
OpenAI's investigation revealed that the campaign "does not appear to have achieved meaningful audience engagement." Most of the social media posts identified by OpenAI received minimal interaction, with a "few or no likes, shares, or comments." The operation also produced content for fake conservative and progressive news outlets, targeting contrasting viewpoints.
Influence campaign's narrative and future concerns
The content generated by the influence campaign suggested narratives such as Donald Trump "being censored on social media and was prepared to declare himself king of the US." Another narrative portrayed Kamala Harris's selection of Tim Walz as her running mate as a "calculated choice for unity." With the US presidential election approaching, there are concerns about potential attempts to manipulate its outcome.
Recent hacking incidents linked to Iran
Last week, former President Trump confirmed his campaign was hacked and linked the incident to a phishing email sent by an Iranian hacking group. In June, the FBI initiated an investigation into the alleged hack on Trump's campaign as well as potential hacking attempts on the Biden-Harris campaign. Earlier this week, it was disclosed that Iranian hackers had targeted both Harris's and Trump's campaigns.
Campaign's threat level and diverse content
OpenAI assessed the threat level of the influence campaign using the Brookings Institution's Breakout Scale. The operation only scored a Category 2 rating (on a scale of one to six), indicating "activity on multiple platforms, but no evidence that real people picked up or widely shared their content." The campaign also generated content on less politically charged topics like fashion and beauty, possibly to appear more authentic or attract followers.
