Introduction

Three senior OpenAI officials responsible for national security policy recently took part in a video call they believed was with a high-ranking Ukrainian representative. In reality, they were speaking with a Russian comedy duo known for prank calls. The conversation touched on military artificial intelligence, Ukraine's defense, and the company's ties to the Pentagon.

What Happened

The pranksters, identified as the Russian duo Vovan and Lexus, arranged the meeting by pretending to be Rustem Umerov, whom they described as the head of Ukraine's Foreign Intelligence Service. According to a recording posted online, the 17-minute discussion included Sasha Baker, who leads national security policy at OpenAI; Caroline Zier, the company's strategic initiatives lead; and Ben Crampton, who oversees European national security matters.

During the call, the executives discussed OpenAI's military support for Ukraine, including the potential provision of offensive AI capabilities. Crampton reportedly said the company wants a deep relationship with Kyiv but may not be able to move immediately on offensive uses. He also noted internal reluctance to involve OpenAI technology in offensive military contexts, while arguing that the line between offense and defense can blur.

Baker mentioned that OpenAI has been in contact with the Pentagon about assisting Ukraine and referenced a $200 million contract aimed at military applications. She also spoke about the company's Daybreak Blue model, designed to patch software vulnerabilities, and an emerging partnership with Ukraine's Ministry of Defense. When the conversation turned to Elon Musk, Baker said OpenAI could not help with him, adding that Musk is not a fan of OpenAI's CEO.

OpenAI later confirmed it investigated the call and determined it was a prank by individuals with a history of targeting officials and companies. The company said it acted internally and with relevant authorities to prevent future incidents.

Why This Matters

The incident exposes gaps in how even sophisticated tech firms verify the identities of foreign officials. It also reveals the extent of OpenAI's engagement with military and intelligence entities, including work tied to the U.S. Department of Defense and Ukraine's armed forces. For observers of AI safety and national security, the call raises questions about how much influence private AI companies should have in wartime and how easily their executives can be manipulated.

Key Takeaways

  • Russian pranksters Vovan and Lexus tricked three OpenAI national security executives into a video call.
  • The impostors claimed to be Rustem Umerov, a senior Ukrainian intelligence official.
  • The 17-minute conversation covered military AI support, offensive capabilities, and OpenAI's Pentagon contract.
  • Executives discussed the Daybreak Blue vulnerability model and a budding partnership with Ukraine's Ministry of Defense.
  • OpenAI confirmed the call was a prank and said it took steps to prevent similar occurrences.

Conclusion

OpenAI's experience serves as a cautionary tale for AI companies operating at the intersection of technology and geopolitics. As artificial intelligence becomes more deeply embedded in military planning, verifying who is on the other end of a video call is no longer optional; it is a core security requirement.