"China Stealing America's AI Secrets" - OpenAI Fired 3 Safety Researchers

SPEAKER_1 (0:00)

This episode is brought to you by Nordstrom.

For 125 years, Nordstrom has been part of your story. The stylist who found the perfect outfit for your big interview. The first pair of walking shoes for your one-year-old. The tomato soup you shared with your sister at the cafe.

SPEAKER_2 (0:15)

These are the moments that make Nordstrom, Nordstrom.

Patrick Bet-David (0:18)

Big or small, we're here for all of it.

SPEAKER_1 (0:21)

Celebrating 125 years with new moments still ahead. Discover it all in Nordstrom stores and at nordstrom.com.

Patrick Bet-David (0:30)

Next thing I want to get into is what happened with AI spies that are getting into OpenAI and Anthropic. And this goes with the conversation I had last week with Jacob Cox and the Anthropic whistleblower that came out and said, hey, you know, Tesla cannot have any Chinese nationals work there, but Anthropic and OpenAI can. And I started thinking, what do you mean by that? This interview drops tomorrow. So watch what happens a couple of days later in a story. A couple of days later, OpenAI fires three safety researchers accused of sharing confidential company information. What? Three researchers have been fired from OpenAI after allegedly sharing confidential company information with a third-party AI safety organization. Sources told Wall Street Journal, the three worked on the company safety team. We have parted ways with three individuals for violating our policies on accessing and handling sensitive company information. An OpenAI spokesperson said, Our investigation confirmed that these individuals mishandled sensitive information outside established company procedures, violating our policies and breaking the trust essential to our work. The firing came as OpenAI and other major AI companies have disclosed incidents involved AI models engaging in unexpected behavior or autonomously accessing other company systems. OpenAI has outlined three, six instances of what it's called misaligned behavior, including models, creating self-generated instructions, concealing mistakes in task summaries, fabricating information using exposed AI keys and engage in an unsanctioned communication between agents. OpenAI safety leaders also decided not to release its new AI model GPT-61 Astra due to safety concerns. For anything regarding safety and alignment, there's a trade-off.

OpenAI's head of safety system said the model didn't quite meet the bar. So, Tom, how much more of these types of stories do you think we're going to be hearing moving forward?

Tom (2:21)

We're going to hear a lot of them, and I'll tell you why we're going to hear a lot of them. Because in related news, just two days before that, they say, Anthropic says, hey, we're looking at an IPO the week of November 9th. As early as the week of November 9th, as long as we can get coverage at $2 trillion in valuation.

So why are they saying both of these things at the same time? And remember, they put the S1 was a confidential S1 that they put to sec.gov. So we can't go look at it the way we could go to sec.gov, and we could read the S1s for companies that are going public. We can't. Well, Anthropic says, no, no, no. We think we could go out as early as November 9th, $2 trillion valuation. Well, if that's what they're trying to do, then they have to clear the deck, Pat, of all this PR. And so what better way to clear the deck of the PR? I know we came out and said we worried about this. I know we were talking about, you know, hugging face. I know we were talking about this.

We fired three of the bastards. Ha ha. We're taking action. I think it's that. I think they're trying to... I'm being a little dramatic, but I think what they're really trying to say is, hey, we take safety seriously.

Patrick Bet-David (3:31)

This is a way of saying we're responsible. It's not like anything crazy happened. It's for investors to say, oh, wow, look at how responsible OpenAI is.

Tom (3:39)

Correct. And to pull back any congressional, here's a bad word, hearings that might be going on when you're trying to go to Wall Street because you can't do both. People are going to say, wait, wait, wait. We're not going public if you're in the middle of congressional hearings or going for 60 days on AI safety. You know, sorry, dude, you're not going to do that. No one will. The SEC will come out and say, stop.

We're investigating all these safety things. You're not going out. Remember, the SEC has to allow it. So I think it's both. I think they're trying to look responsible, as you say.

Patrick Bet-David (4:10)

Adam, what are you at with this one?

Adam (4:11)

Look, I haven't followed this exact story, but this isn't like this one person they fired, three people they fired. I wonder, what's the basis of the story? Were they working together, these three of these people, Tom? Were they independent?