Russian pranksters tricked three OpenAI natsec and policy executives into a video meeting by posing as Ukrainian officials; OpenAI says it "acted internally"
First reported by Theintercept ·
Your personal data shared in professional contexts may be compromised by sophisticated impersonation tactics.
Russian pranksters, known as Vovan and Lexus, deceived three senior OpenAI national security and policy executives into a 17-minute video call by impersonating Ukrainian officials. The executives, Sasha Baker, Caroline Zier, and Ben Crampton, discussed OpenAI's military support for Ukraine, including the potential for providing offensive AI capabilities. The pranksters posed as Rustem Umerov, head of Ukraine's Foreign Intelligence Service. During the call, OpenAI officials indicated a desire for a deep relationship with Ukraine and acknowledged some reluctance from senior staff regarding offensive military applications of their technology. They also mentioned ongoing work with the U.S. Department of Defense and a nascent partnership with the Ukrainian Ministry of Defense. OpenAI stated it has conducted an internal investigation and taken action to prevent future incidents.
This incident highlights significant vulnerabilities in how technology companies handle sensitive discussions with foreign governments, particularly concerning national security and military applications. The ease with which the pranksters bypassed OpenAI's security protocols for a high-level discussion suggests a need for more robust verification and vetting processes for virtual meetings, especially those involving international relations and defense.
The conversation's content, touching on offensive AI capabilities and existing military contracts, underscores the growing intersection of AI development and geopolitical conflict. It also reveals internal discussions at OpenAI regarding the ethical boundaries of deploying their technology in warfare, indicating a tension between company policy and the urgent demands of ongoing conflicts.
AI-written summary. May contain errors.