Attempts to manipulate social media through the use of ChatGPT and other OpenAI tools have largely been unsuccessful, according to a recent report by OpenAI. While these tools can facilitate the creation of bios and social media posts, helping to scale spam networks without needing large teams, their over-reliance on AI has proven to be a vulnerability rather than a strength.
OpenAI highlighted a particular election interference case over the summer that was swiftly neutralized, predominantly due to the attackers’ dependence on AI tools. OpenAI noted that disrupting the AI-driven processes halted the offending social media operations during crucial election periods.
Despite some deceptive campaigns managing to engage audiences online, OpenAI concluded that its tools have not significantly advanced the capabilities of threat actors to create new, disruptive content. What these tools do offer are incremental capabilities similar to those available in non-AI powered tools.
OpenAI’s findings underscore the need for collaborative efforts in building robust defenses against cyber threats and influence operations involving state-linked actors. Through improved detection, AI companies are hopeful to identify connections between disparate sets of threat activities, potentially strengthening the internet’s overall defense framework.
The data from OpenAI illustrates an acknowledged potential in AI models for analyzing malicious attachments, as in phishing campaigns, a prospect suggesting progress in the AI industry may continue to disrupt cyber threats effectively.