OpenAI publishes 'Influence and cyber operations: an update' (October 2024)
OpenAI reported disrupting Russian ('Stop News') and Iranian ('Storm-2035') influence operations targeting elections, alongside continued state-linked misuse for coding and translation, with no evidence of novel capability gains.
- Security & misuse
- Minor
OpenAI published its third recurring threat-intelligence report, describing more than 20 state-linked and other malicious networks it had disrupted since the start of 2024 for misusing its models.
Named operations included “Stop News,” a Russia-linked network generating English- and French-language content and DALL-E images for pro-Russian sites; the Iranian “Storm-2035” operation, which OpenAI had banned months earlier for producing partisan US election content aimed at Latino and Venezuelan audiences; and separate Iranian and Chinese state-linked actors using the models for malware debugging, reconnaissance and translating material into Persian. None of the disruptions were novel capability discoveries — the uses documented were content generation, debugging and translation, tasks the actors could already attempt by other means.
We have not seen evidence of this leading to meaningful breakthroughs in their ability to create substantially new malware or build viral audiences.
That conclusion — misuse continuing, but no evidence of AI conferring a decisive new capability — was consistent with OpenAI’s earlier reports from February and May 2024 and with similar findings from other labs. It became a recurring reference point in the debate over whether frontier models were meaningfully lowering the barrier to cyber-offence or influence operations, as opposed to merely making existing low-grade activity somewhat cheaper.