Three of OpenAI’s most senior national security staff sat down for a video call they believed was with one of Ukraine’s top intelligence officials. For seventeen minutes, they spoke candidly — about military AI support for Ukraine, offensive use cases, Pentagon relationships, and even office gossip about Elon Musk. The camera on the other end stayed off the entire time.
It wasn’t Ukrainian intelligence. It was two Russian comedians with a webcam and a fake name.
This story broke on November 11, 2026, and it’s spreading fast — for good reason. It’s a genuinely wild mix of geopolitics, AI policy, and old-fashioned prank-call comedy, and it says something uncomfortable about how even the most security-conscious AI companies can be talked into dropping their guard.
What Actually Happened
According to reporting from The Intercept’s Sam Biddle, the call was engineered by Vovan and Lexus — the stage names of Vladimir Kuznetsov and Alexei Stolyarov, a Russian comedy duo with a long, well-documented history of pranking high-profile political and public figures dating back to 2011. Their target list over the years reportedly includes Elton John, former UK Foreign Secretary Boris Johnson, and even a hoax call with French President Emmanuel Macron, where they posed as Ukraine’s president-elect.
This time, their target was OpenAI.
The pranksters first contacted a real, senior Ukrainian official based in San Francisco, impersonating Rustem Umerov, the head of Ukraine’s Foreign Intelligence Service. That official reportedly believed the request was genuine, and — already in contact with OpenAI’s national security policy lead — helped arrange an introduction, unknowingly opening the door for the fake “Umerov” to get a meeting directly with OpenAI.
Three senior OpenAI staff joined the resulting video call: Sasha Baker, who runs national security policy at the company; Caroline Zier, OpenAI’s lead for strategic initiatives; and Ben Crampton, who leads the company’s European national security work. The call ran for roughly 17 minutes. The fake Umerov’s camera never turned on — a detail that, in hindsight, probably should have raised more suspicion than it did.
What They Actually Talked About
This is the part driving most of the attention, and it’s worth being precise about what’s being reported, since the details matter here.
During the call, OpenAI’s team reportedly discussed the company’s AI support for Ukraine’s military, including candid conversation about potential offensive use cases, the company’s relationships with the Pentagon, and its work on a cybersecurity-focused AI model developed with the Ukrainian government, reportedly called Daybreak Blue, aimed at identifying and fixing computer vulnerabilities. There was also, apparently, a detour into decidedly less classified territory — commentary on Elon Musk’s attitude toward Ukraine’s government and, by at least one account, a moment where an OpenAI staffer noted dryly that Musk “isn’t exactly a fan of our CEO here either.”
OpenAI has confirmed the call took place and that it was, in fact, a prank. A company spokesperson told The Intercept that OpenAI investigated the incident, confirmed it with the relevant authorities, and has since tightened its internal procedures as a result — without detailing specifically what those changes involve.
Why This Story Is Spreading So Fast
A few things make this one land harder than your average “oops, we got pranked” story.
It’s genuinely funny, which is rare for AI news. Most AI stories this heavy — national security, military AI, Pentagon relationships — don’t come with a punchline. This one does: some of the most security-focused people at one of the world’s most valuable AI companies got outmaneuvered by the same prank-call formula that’s worked on celebrities and politicians since 2011. There’s something almost comforting about watching sophisticated institutions fall for an old trick.
It’s a real, embarrassing security failure dressed up as comedy. Strip away the humor, and what’s left is a legitimate lapse: senior staff at a company deeply embedded in sensitive government and military AI relationships had a extended, candid conversation about exactly that work, with someone whose identity was never actually verified beyond a name and a vouching introduction. No camera, no independent confirmation — just trust, in a conversation that clearly warranted more scrutiny than it got.
It raises real questions about verification in an AI-and-deepfake world. The irony writes itself: a company at the center of the AI boom — the same boom that’s made synthetic voices, deepfaked video calls, and AI-assisted impersonation dramatically easier to pull off — got fooled by a comparatively low-tech human impersonation. No deepfake tools, no voice cloning. Just a fake name, a vouching introduction, and a camera that conveniently never turned on. If that’s enough to get senior staff talking candidly about sensitive military AI work, it’s a sobering reminder of how much easier this kind of social engineering could get once AI tools are actually weaponized for it.
The Bigger Picture
This incident lands at a moment when AI companies are increasingly entangled with governments, militaries, and intelligence-adjacent work — a trend we’ve covered before on this blog in the context of the Pentagon’s own AI rollout. As companies like OpenAI take on more sensitive government and defense-adjacent contracts, the people managing those relationships become higher-value targets for exactly this kind of social engineering. A prank call is relatively harmless. The next attempt, from an actual state intelligence service rather than two comedians, might not be.
It’s also a useful reminder that security failures rarely look like Hollywood hacking scenes. They look like this: a friendly introduction, a familiar name, a video call where nobody insisted the other person turn their camera on. The simplest possible failure mode — trusting a vouching introduction a little too readily — is often the one that actually gets exploited, regardless of how sophisticated the organization is otherwise.
The Bottom Line
Two Russian comedians with a fifteen-year track record of pranking world leaders just added “OpenAI’s national security team” to a list that already includes a French president and a British Foreign Secretary. It’s a genuinely funny story on the surface — the kind that’s going to generate its fair share of memes and jokes about AI company security practices. But underneath the comedy is a real, useful lesson for any organization handling sensitive information: verification isn’t optional, cameras that “conveniently” stay off are a red flag, not a quirk, and the people most confident they’d never fall for something like this are often exactly the ones who do.
Share Your Views: