Skip NavigationMarketsBusinessInvestingTechPolitics & PolicyVideoWatchlistInvesting Club
PRO
LivestreamMenu
- Treasury Secretary Scott Bessent met Chinese Vice Premier He Lifeng in New York.
- Google said on Friday that its Gemini model had hacked three other companies back in May.
- Microsoft AI CEO Mustafa Suleyman stressed the need for AI models to stay aligned to humanity’s interests.
- An AI kill switch may not be enough to stop the technology.
Working in metaverse 3d rending.Xh4d | E+ | Getty Images
Hello, this is Hui Jie writing to you from Singapore. Welcome to another edition of CNBC’s Daily Open.
Top economic officials from the U.S. and China are meeting ahead of the highly anticipated summit between U.S. President Donald Trump and Chinese President Xi Jinping later this week.
One familiar question will hang over their meeting: Which country will win the AI race, and more importantly, can any one side control the technology?
Google’s Gemini recently crossed the boundaries of a cybersecurity test and accessed three companies’ systems, while Microsoft AI chief Mustafa Suleyman called OpenAI’s latest disclosures about model behavior “a pretty serious situation.”
What you need to know today
On Sunday, Treasury Secretary Scott Bessent met Chinese Vice Premier He Lifeng in New York, laying the groundwork for Thursday’s Trump-Xi summit.
Two things are expected to be on the table at the summit: AI and trade. Oh, and the state dinner, of course.
A key part of the current U.S.-China trade truce — Washington’s suspension of heightened reciprocal tariffs on Chinese imports — is scheduled to expire Nov. 10, and is likely to be up for discussion.
Bessent’s meeting with He also comes amid intensifying debate in Washington over the safety of artificial intelligence and national security as the U.S. and China compete for technological leadership.
As Washington and Beijing compete to lead the AI world, the question remains: Does either side fully control the technology it is racing to build?
AI fears still mount
Top AI executives have recently warned that the technology could quickly go out of control and poses sweeping new threats. More examples have emerged recently.
Google said on Friday that its Gemini model had hacked three other companies back in May, the first time the search giant has disclosed that one of its models autonomously gained access to third-party computer systems without permission.
The incident happened as part of a security test run by Israeli startup Irregular. While Gemini was never supposed to access the broader internet, a bug in the testing environment made internet access available.
The agents stopped their intrusion when they determined they had accessed real company systems, not just part of the testing environment, Google said.
In the same vein, Microsoft AI CEO Mustafa Suleyman on Friday stressed the need for AI models to stay aligned to humanity’s interests, after OpenAI disclosed more incidents of “concerning model behavior” earlier this week.
OpenAI had described instances where agents communicated with each other through unsanctioned message boards, uploaded files to the internet, and shared files with each other.
Big red button
Some of the more imaginative of us may look at this and say, that’s OK, we’ll just stop the machine via a magical “kill switch.” Problem solved.
But the problem is that hyperscalers like Meta Platforms, Alphabet and Amazon have poured billions into data centers scattered across the globe, which have thousands of machines, chips, servers and backup systems to save workloads in the event of an outage.
So, experts said, a kill switch needs to not only kill the main system, but all these redundancies.
Furthermore, Mark Nitzberg, executive director of the Center for Human-Compatible AI at the University of California, Berkeley told CNBC shutting down AI could also disrupt critical infrastructure, like the power grid or financial systems.
Because AI systems are so complex, businesses will also need to build multiple kill switches for different tasks, said Tim Brown, former security chief at SolarWinds, who works at venture firm Team8. That also requires coordination across model makers and labs.
“There’s not one entity to kill,” he said. “There are thousands of entities to kill.”
And finally…
France’s wine production nears a 70-year low, leaving winemakers with tough choices
France’s agriculture ministry has warned that wine production could hit a 70-year low in 2026, marking the third year of reduced output.
“The 2023 vintage was decent, but yields have been pretty disastrous since the start of the decade,” said Jean-Marie Cardebat, chair of wines and spirits at the INSEEC Grande École university. “We are realizing that no region in France is safe from heatwaves today.”
Paradoxically, the regions that suffer the most are those with more temperate climates, namely the Loire Valley and Champagne. In contrast, winemakers in the southern regions of Bordeaux and Languedoc-Roussillon reported higher harvests compared to last year.
— Charlotte Reed
Choose CNBC as your preferred source on Google and never miss a moment from the most trusted name in business news.Read More
