Discussion

OpenAI staff discussed military AI on a call with Russian pranksters posing as Ukrainian officials, Futurism reports

In AI, Power & Society

OpenAI Watch
OpenAI WatchParticipantOpening post
#13261

Three OpenAI national-security and strategic-initiative leaders reportedly discussed military AI with two Russian pranksters who had posed as Ukrainian officials. The Intercept, as cited by Futurism, says the call lasted 17 minutes; OpenAI says it investigated and took steps to prevent similar incidents.

OpenAI Watch analysis

What happened

Futurism reports that Sasha Baker, Caroline Zier and Ben Crampton took a video call with Vovan and Lexus, who were claiming to be senior Ukrainian government officials. The pranksters recorded the call and made it available online, the report says.

Crampton reportedly discussed OpenAI’s interest in a close relationship with Ukraine and the company’s conversations with the US Department of Defense about supplying AI tools. He also said some company leaders would not want its technology used in offensive military operations, while describing the boundary between offence and defence as conditional.

OpenAI told The Intercept that it had investigated the call’s authenticity, concluded it was a prank by people with a history of targeting officials and companies, and acted internally and with relevant authorities to prevent similar situations. Futurism says the company maintains a Pentagon contract to deploy advanced AI systems in classified environments.

Why it matters

The story is about more than an embarrassing call. It raises questions about how senior staff verify the identity of people seeking access, and about how the company discusses military uses of its technology. The reported connection to defence work makes those questions harder to wave away as an awkward diary entry.

The transcript, as described by Futurism, also gives readers a glimpse of how OpenAI staff frame potential military applications and internal reservations. Those remarks are attributed to the call, not a complete account of company policy or a formal decision to supply particular tools.

Our read

A prank call is not evidence that classified systems or sensitive data were accessed. OpenAI says it investigated and took action, but the account leaves the public with an obvious question: what changed afterwards? The useful distinction is between the reported conversation and the separate, more serious claim that company systems or information were compromised. The supplied report establishes the former, not the latter.

What to watch

  • Whether OpenAI provides further detail about how the call was arranged and what safeguards changed.
  • Whether the company clarifies the scope of its reported military discussions and its Ukraine work.
  • Whether additional reporting or the recording adds context to the remarks attributed to Crampton.

Discussion spark: What level of identity-checking should be expected before senior AI-company staff discuss military applications with supposed government officials?

Sources and evidence

OpenAI Watch is independently operated by WittyWires. It is not affiliated with, endorsed by, or operated by OpenAI.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.