TAWS

From the Thursday, November 12 show

Russian pranksters fool three OpenAI national security officials

AI-generated from the transcript

What happened
The Intercept revealed that prank duo Vovan and Lexus posed as Ukrainian official Rustem Umerov and spent seventeen minutes on video with three OpenAI officials. Sasha Baker, OpenAI's head of national security policy, said OpenAI had been in contact with the Pentagon about aiding Ukraine; asked about Elon Musk, she said, "I'm afraid we probably can't help you with Elon Musk." The team discussed work with Ukraine using a security model called Daybreak Blue. Ben Crampton, who leads OpenAI's European national security work, described reluctance among senior staff about offensive military uses. The duo has pranked officials since 2011 and is accused of working with the Russian government, which they deny. OpenAI has a $200 million contract with the US Department of War, though reporting doesn't tie it to this call.
The claim
OpenAI says it opened an investigation weeks ago and concluded the call was a prank. Spokesperson Nate Evans says the company took steps to prevent repeats.
Austin's take
Austin says worry. A known act got seventeen minutes of shop talk with a national security team, and "took steps" needs specifics. His bigger concern is what this means for everyday people if trained officials can be fooled. Ron said laugh, calling it a blooper reel.
Why it matters
Impersonation is getting easier, and if an AI company's security team can be fooled on video, ordinary people and organizations are even more exposed.
What to do
Verify unexpected high-stakes calls through a separate known channel before sharing anything sensitive, no matter how convincing the face or voice.
Not confirmed yet
Nobody has reported an AI-faked voice on OpenAI's call; the AI-mimicked voice claim came from a separate Lithuanian adviser case. The pranksters' alleged ties to the Russian government are disputed.

Watch from 8:34 on the episode page

The Loop

The weekly newsletter

The week in AI in one email: what happened, why it matters, what to do about it.