A New Chinese AI Resets the Global Tech Race

Jack Gillam (0:00)

This podcast is brought to you by Federated Hermes. We put our investments through a ruthless vetting process because we don't like surprises and neither do our clients.

SPEAKER_2 (0:08)

Learn more at federatedhermes.com/us.

Jack Gillam (0:12)

Investments are subject to risk and may lose value.

Luke Vargas (0:17)

A big leap for Chinese AI as researchers find that a new model goes toe-to-toe with Anthropic in cybersecurity. Plus, we'll look at how your unsecured smart devices could be playing a role in massive cyberattacks.

Kareem Maruz (0:32)

What we're seeing is through these ResProxy networks, that's the initial entry point that creates a whole set of computers, laptops, streaming devices, picture frames, smart refrigerators, all hitting the same site at the same time. Very few sites in the world are resilient to that level of inbound traffic.

Luke Vargas (0:51)

And a staggering death toll starts coming into focus after last week's Venezuela earthquakes. It's Monday, June 29th. I'm Luke Vargas for the Wall Street Journal, and here is the AM Edition of What's News, the top headlines and business stories moving your world today.

China has caught up with Anthropic in cybersecurity, potentially resetting the global AI race. Security researchers say that a new model released by China's ZAI lags behind products from Anthropic and OpenAI in some tasks, but can match them in finding security bugs. The release of the model comes as US lawmakers have tried restricting access to the most potentially disruptive AI models, like Anthropic's Mythos. However, as journal tech reporter Sam Schechner explained, Z.AI's model, being open-weight, meaning it can be downloaded, used, and modified without supervision, makes that impossible.

Sam Schechner (1:50)

What that means is that there's a kind of toothpaste out of the tube scenario. You have on one hand the US trying to lock down some of these capabilities. I mean, you have Anthropic saying, okay, we're not going to release the most powerful version of Mythos beyond a certain set of trusted partners. And then you even have the administration telling Anthropic that it had to pull back its fable model because the guard rails against foreign use weren't high enough. So the question of whether or not these capabilities that could help enable potential cyber attacks are going to get out there is becoming a much more real one.

And then on the other side of the street practically, you have another company giving away models with similar capabilities for free. And so if those models are out there and they have similar capabilities, then there's some critics who say the US is just shooting itself in the foot by not allowing US companies to sell and distribute their models as well.

Luke Vargas (2:46)

Critics of the White House AI strategy say it's been lax in restricting use of Chinese open-weight models which are popular among US businesses. Jacob Helberg, Undersecretary of State for Economic Affairs, defended the administration, saying it's quote, very much focused on Chinese open-source models.

South Korean stocks have pared their losses today after the country's president announced a plan to invest nearly $2 trillion in the country's semiconductor sector, data centers, and robotics. Memory chip giants Samsung and SK Hynex also committed to substantial new investments to build chip-making facilities. SK Hynex, Samsung, and US company Micron Technology are the world's leading suppliers of high-bandwidth memory chips which are crucial components for NVIDIA's AI accelerators. Shares of both South Korean companies have been on a wild ride this year on the back of the AI buildout, with both crossing the $1 trillion market value threshold.

After days of back and forth fighting, the US and Iran have agreed to resume peace talks. The latest flare-up began on Thursday and reached a peak over the weekend as the US hit a range of Iranian targets, and Tehran said that it targeted American bases in Kuwait and Bahrain. Middle East correspondent Benoit Foucault told me that after things potentially heading towards a return to war, talks in Qatar are reportedly back on.

Benoit Foucault (4:13)

They would take place possibly this week in Doha, which is one of the parties that was attacked by Iran, but it's the Kim mediator. We will focus on the threat of homers, trying to find an understanding between both sides. But potentially, I would suspect also on the release of cash that is held in Qatar. And most importantly, both sides agrees to stop attacking each other.

Luke Vargas (4:33)

Benoit, is there any real appetite, though, for potentially releasing money to Iran just in the last few days, the country has been launching new attacks around the region?

Benoit Foucault (4:42)

First of all, remember, we are talking about a cycle where there's always strike and talk. There's always attacks, and then there's negotiation, there's attacks, and there's negotiations again. So obviously, the question would be whether, like you say, there would be an agreement to release some cash when there could be a suspicion that it could be used to fund the military. Now on that, there is a limitation. I mean, the US has said it's only humanitarian money, only for food, so that benefits really the Iranian people.