Connect with us

Published

on

Voters cast their mail-in ballots at a ballot drop box outside Maricopa County Recorder and Elections Department southeast Mesa office during the Arizona state primary election in Mesa, Arizona, U.S. July 30, 2024. 

Rebecca Noble | Reuters

Derek Bowens has never had such an important job. He’s the director of elections in Durham County, North Carolina, one of the most-populous areas of a state that’s increasingly viewed as crucial to the 2024 presidential contest.

So when a former precinct official emailed Bowens in July to warn him of a post containing voting misinformation that was spreading virally on Facebook, Bowens quickly recognized that he may be facing a crisis.

The post, written as if from an authority on the subject, said voters should request new ballots if a poll worker, or anyone else, writes anything on their form, because it would be invalidated. The same incorrect message was spread on Facebook during the 2020 election, but the platform flagged the content at the time as “false information” and linked to a story that debunked the rumor by Facebook’s fact-checking partner, USA Today.

Bowens said no such tag appeared on the post, which was widespread enough that the North Carolina State Board of Elections had to issue a press release on Aug. 2, informing voters that false “posts have been circulating for years and have resurfaced recently in many N.C. counties.”

“It was spreading and there wasn’t anything happening to stop it until our state put out a press release and we started engaging with our constituency on it,” Bowens told CNBC in an interview.

The elections board wrote a post on Facebook, telling voters to “steer clear of false and misleading information about elections,” with a link to its website. As of Wednesday, the post had eight comments and 50 shares. Meanwhile, multiple Facebook users in states like North Carolina, Mississippi and New Jersey continue to share the ballot misinformation without any notification that it’s false.

CNBC flagged posts with the false information to Meta. A company spokesperson said, “Meta has sent them to third-party fact-checkers for further review.”

Across the U.S., with 40 days until the Nov. 5 election, state and local officials say they are puzzled by what to expect from Facebook. Like in the past two presidential election cycles, the spread of misinformation on the social network has threatened to disrupt voting in what’s expected to be another razor-thin contest decided by thousands of voters in a handful of states. Recently, a Facebook post containing a false claim about Haitian immigrants eating pets in Springfield, Ohio, ballooned out of control and gained resonance after it was repeated by Republican nominee Donald Trump in a debate.

In 2016, Facebook was hammered by Russian operatives, pushing out false posts about Hillary Clinton to bolster Trump. In 2020, the site hosted rampant misinformation about politically charged issues like Covid treatments, masking and voter fraud.

The big difference this go-round is that Facebook has largely removed itself from the equation. In 2021, Meta began pushing political and civic content lower in its algorithms, which contributed to a dramatic decline in news traffic last year for publishers. Earlier this year, Meta announced that it would deprioritize the recommendation of political content on Instagram and its Twitter-like Threads service, a move the company said more aligns with what consumers want to see on their feeds.

Still, posts with false information can spread rapidly across wide swaths of users along with comments that amplify the misinformation, and government agencies have little ability to counteract them, because they have such limited reach on the platform.

New report shows AI-powered chatbots could spread election misinformation

And while Facebook has lost some of its prominence due in part to the rise of TikTok, particularly among younger audiences, the site still had more than 200 million daily users in the U.S. and Canada at the end of last year, the last time it issued regional numbers. Facebook and Instagram are generally both in the top 10 among the most-visited websites and most-popular apps in the U.S, according to the Pew Research Center and Similarweb.

Interviews with nearly a dozen regional and statewide government officials with election-related duties reveal the challenges they say they’re having using and monitoring Meta’s apps, as well as other social networking services like X, now owned by Elon Musk. The officials say they’re working overtime to ensure the safety and integrity of the election but say they’re receiving little effective help from the companies, which scaled back their trust and safety teams as part of broader cost-cutting efforts that began in 2022.

Meta ultimately cut 21,000 jobs, including in trust and safety and customer service, over multiple rounds of layoffs. As CNBC reported last year, the company dissolved a fact-checking tool that would have let news services like The Associated Press and Reuters, as well as credible experts, add comments at the top of questionable articles as a way to verify their trustworthiness. Reuters is still listed as a fact-checking partner, but an AP spokesperson said the news agency’s “fact-checking agreement with Meta ended back in January.” 

The Meta spokesperson told CNBC in a statement that the company’s “integrity efforts continue to lead the industry and we have around 40,000 people globally working on safety and security — more than we had during the 2020 cycle.” The company says it now partners with about 100 third-party fact-checking groups across the globe “who review and rate viral misinformation in more than 60 languages.”

Challenges in Maricopa County

Like North Carolina, Arizona is one of the seven swing states expected to determine whether Trump or Vice President Kamala Harris, the Democratic nominee, win the presidency.

That reality has put Taylor Kinnerup in the spotlight. Kinnerup is the communications director for the recorder’s office of Maricopa County, home to more than half of Arizona’s population.

Kinnerup and her colleagues use social media to distribute up-to-date information about election-related procedures, like when residents can mail in early ballots or where to find their voting center. It’s a particularly sensitive job following Trump’s false claims of voter fraud in Arizona in 2020, when the state went blue for the first time in a presidential contest since 1996.

Given Maricopa County’s high profile during the election season, the state often attracts attention from Facebook users across the country. Many of them, Kinnerup said, are older and still leave comments about debunked conspiracy theories, such as the false claim that Sharpie markers invalidate ballots.

Kinnerup said her team places “extreme emphasis on constant communication and transparency to the public,” actively sharing election-related content across Facebook and Instagram, particularly during peak hours when it’s more likely to reach voters.

A few months ago, Kinnerup discovered that her office’s Facebook and Instagram accounts were no longer linked, meaning she couldn’t access the apps using the same credentials, or automatically schedule a single post to go across both sites.

Ahead of the primary elections in July, Kinnerup said she struggled to resolve the account issues with Meta. She said she engaged in a monthslong email exchange with numerous representatives, but found there was “no way to really make progress.” When she did get a response, it was little more than a canned statement, Kinnerup said.

Meanwhile, Kinnerup is busy overseeing media and constituent tours of the county’s election facilities to help dispel false notions that the process is being rigged as her office continues to deal with the fallout of the 2020 election. Kinnerup said she led more than 20 such tours in June.

“I couldn’t be dealing with Meta every single day, because I had to be giving tours,” Kinnerup said. The time spent trying to find a fix “was a huge issue for me,” she said.

By the time Kinnerup said she’d resolved her account issues, in mid-July, she and her colleagues had wasted countless hours on the problem, leaving her team to “feel we were put in a position where the full message we were trying to get out wasn’t ever fully there.”

Even with her office’s Facebook and Instagram accounts working again, Kinnerup says their organic social media posts generate little engagement, and her team has used sponsored ads to help expand reach across the platforms. Her team has continued with the facility tours, leading 25 this month.

Meta’s spokesperson said the company has been hosting training sessions for state and local officials since February, informing them of tools like voting alerts, which allow them to send messages to people in their area.

Former US President and Republican presidential candidate Donald Trump leaves at the end of a presidential debate with US Vice President and Democratic presidential candidate Kamala Harris at the National Constitution Center in Philadelphia, Pennsylvania, on September 10, 2024. 

Saul Loeb | AFP | Getty Images

“There are multiple channels by which officials can reach us, including teams responsible for specific states and regions, and our ability to respond to them remains unchanged,” the spokesperson said.

Kinnerup said she was not “aware of any of this,” and in her year in the role has “never received any direct communication with Meta that I’m aware of.”

Bowens told CNBC in a follow-up email that he “was not aware of the sessions or the tools.”

Congress is well aware of potential problems. During a Senate hearing last week on election threats, Meta’s head of global affairs, Nick Clegg, fielded questions about the company’s election preparedness. Sen. Susan Collins, R-Maine, expressed concern about the safety and integrity of “down-ballot races at the state level, county level, local level.”

Intelligence agencies, Collins said, have told senators that bad actors from China could be focusing on disrupting regional races as opposed to the presidential election, and that state and regional officials “are far less likely to receive the kinds of briefings that we receive or to get information from Homeland Security or the FBI on how to be on alert.”

Clegg said Collins was “right to be concerned” and that Meta’s “vigilance needs to be constant.”

“It can’t just sort of peak at the time of the presidential elections,” Clegg said.

‘Three people will see it’

For Scott McDonell, the Dane County clerk in the swing state of Wisconsin, it’s been difficult to share accurate voting information on Facebook from his office’s official government account, which only has 608 followers on Facebook. McDonell said his posts get very little traction compared with years past.

“If I link to a story about election security, three people will see it,” McDonell said. Posts that include pictures do marginally better, he said, because “Facebook likes pictures.”

“Don’t link to an article, that will go to zero,” he said.

McDonell said many of his colleagues have “gotten abused” so much on Facebook in recent years that they don’t post about elections anymore.

“Basically, your average county clerk is terrified of it, and they just do it to share baby photos,” McDonell said.

Consternation ahead for equities before November's elections, says Wells Fargo's Sameer Samana

In Los Angeles County, Jeramy Gray, the chief deputy of the registrar-recorder/county clerk office, said small government offices often lack the resources needed to effectively utilize social media and to troubleshoot problems.

Meta “recently put a team together to assist” his office, Gray said, adding that the company appears to be the “most mature” of the big platforms even if it’s not a “model partner.”

“What I would like to see is just more engagement from them, at least three to four months from a large national election, for them to reach out to key stakeholders at the state and local level to really talk about what they can do or what they’re doing,” Gray said.

Bowens, in North Carolina’s Durham County, said the tech platforms could be much more helpful in assisting his office and others as they navigate through some of the confusion about what type of content is acceptable.

Bowens said he’s concerned about acting too aggressively because of potential censorship issues and recognizes there’s a gray area between misinformation and citizens exercising their First Amendment rights.

“You know, we’ve got a very diverse election system in this country,” Bowens said. “What was on that post may very well be true in another state. Therefore, is it misinformation?”

WATCH: Sizing up Trump and Harris’ economic plans

Sizing up Trump and Harris' economic plans

Continue Reading

Technology

SoftBank founder Son makes his biggest bet by staking the Japanese giant’s future on AI

Published

on

By

SoftBank founder Son makes his biggest bet by staking the Japanese giant's future on AI

Masayoshi Son, chairman and chief executive officer of SoftBank Group Corp., speaks at the SoftBank World event in Tokyo, Japan, on Wednesday, July 16, 2025.

Kiyoshi Ota | Bloomberg | Getty Images

Masayoshi Son is making his biggest bet yet: that his brainchild SoftBank will be the center of a revolution driven by artificial intelligence.

Son says artificial superintelligence (ASI) — AI that is 10,000 times smarter than humans — will be here in 10 years. It’s a bold call — but perhaps not surprising. He’s made a career out of big plays; notably, one was a $20 million investment into Chinese e-commerce company Alibaba in 2000 that has made billions for SoftBank.

Now, the billionaire is hoping to replicate that success with a series of investments and acquisitions in AI firms that will put SoftBank at the center of a fundamental technological shift.

While Son has been outspoken about his vision over the last year, his thinking precedes much of his recent bullishness, according to two former executives at SoftBank.

“I vividly remember the first time he invited me to his home for dinner and sitting on his porch over a glass of wine, he started talking to me about singularity – the point at which machine intelligence overtakes human intelligence,” Alok Sama, a former finance chief at SoftBank until 2016 and and president until 2019, told CNBC.

SoftBank’s big AI plays

For Son, AI seems personal.

“SoftBank was founded for what purpose? For what purpose was Masa Son born? It may sound strange, but I think I was born to realize ASI,” Son said last year.

That may go some way to explain what has been an aggressive drive over the past few years — but especially the last two — to put SoftBank at the center of the AI story.

In 2016, SoftBank acquired chip designer Arm in a deal worth about $32 billion at the time. Today, Arm is valued at more than $145 billion. While Arm blueprints form the basis of the designs for nearly all the world’s smartphones, these days, the company is looking to position itself as a key player in AI infrastructure. Arm-based chips are part of Nvidia’s systems that go into data centers.

In March, SoftBank also announced plans to acquire another chip designer, Ampere Computing, for $6.5 billion.

ChatGPT maker OpenAI is another marquee investment for SoftBank, with the Japanese giant saying recently that planned investments in the company will reach about 4.8 trillion Japanese yen ($32.7 billion).

SoftBank has also invested in a number of other companies related to AI across its portfolio.

“SoftBank’s AI strategy is comprehensive, spanning the entire AI stack from foundational semiconductors, software, infrastructure, and robotics to cutting-edge cloud services and end applications across critical verticals such as enterprise, education, health, and autonomous systems,” Neil Shah, co-founder at Counterpoint Research, told CNBC.

“Mr. Son’s vision is to cohesively connect and deeply integrate these components, thereby establishing a powerful AI ecosystem designed to maximize long-term value for our shareholders.” 

Stock Chart IconStock chart icon

hide content

SoftBank’s stock performance since 2017, the year that its first Vision Fund was founded.

There is a common theme behind SoftBank’s investments in AI companies that comes directly from Son — namely, that these firms should be using advanced intelligence to be more competitive, successful, to make their product better and their customers happy, a person familiar with the company told CNBC. They could only comment anonymously because of the sensitivity of the matter.

It started with and brain computers and robots

As SoftBank launched “SoftBank’s Next 30-Year Vision” in 2010, Son spoke about “brain computers” during a presentation. He described these computers as systems that could learn and program themselves eventually.

And then came robots. Major tech figures like Nvidia CEO Jensen Huang and Tesla boss Elon Musk are now talking about robotics as a key application of AI — but Son was thinking about this more than a decade ago.

In 2012, SoftBank took a majority stake in a French company called Aldebaran. Two years later, the two companies launched a humanoid robot called Pepper, which they billed as “the world’s first personal robot that can read emotions.”

Later, Son said: “In 30 years, I hope robots will become one of the core businesses in generating profits for the SoftBank group.”

SoftBank’s bet on Pepper ultimately flopped for the company. SoftBank slashed jobs at its robotics unit and stopped producing Pepper in 2020. In 2022, German firm United Robotics Group agreed to acquire Aldebaran from SoftBank.

But Son’s very early interest in robots underscored his curiosity for AI applications of the future.

“He was in very early and he has been thinking about this obsessively for a long time,” Sama, who is author of “The Money Trap,” said.

In the background, Son was cooking up something bigger: a tech fund that would make waves in the investing world. He founded the Vision Fund in 2017 with a massive $100 billion in deployable capital.

SoftBank aggressively invested in companies across the world with some of the biggest bets on ride hailing players like Uber and Chinese firm Didi.

But investments in Chinese technology companies and some bad bets on firms like WeWork soured sentiment for the Vision Fund as it racked up billions of dollars of losses by 2023.

Vision but bad timing

The market questioned some of Son’s investments in companies like Uber and Didi, which were burning through cash at the time and had unclear unit economics.

But even those investments spoke to Son’s AI view, according to the former partner at the SoftBank Vision Fund.

“His thought back then was the first advent of AI would be self-driving cars,” the source told CNBC.

Again this could be seen as a case of being too early. Uber created a driverless car unit only to sell it off. Instead, the company has focused on other self-driving car companies to bring them onto the Uber platform. Even now, driverless cars are not widespread on roads, though commercial services like those of Waymo are available.

SoftBank still has investments in driverless car companies, such as British startup Wayve.

Timing clearly wasn’t on Son’s side. After record losses at the Vision Fund in 2022, Son declared SoftBank would go into “defense” mode, significantly reducing investments and being more prudent. It was at this time that companies like OpenAI were beginning to gain steam, but still before the launch of ChatGPT that would put the company on the map.

“When those companies came to head in 2021, 2022, Masa would have been in a perfect place but he had used all his ammunition on other companies,” the former Vision Fund exec said.

“When they came to age in 21, 22, the Vision Fund had invested in five or six hundred different companies and he was not in a position to invest in AI and he missed that.”

Son himself said this year that SoftBank wanted to invest in OpenAI as early as 2019, but it was Microsoft that ended up becoming the key investor. Fast forward to 2025, the Vision Fund — of which there are now two — has a portfolio stacked full of AI focused companies.

But that period was tough for investors across the board. The Covid-19 pandemic, booming inflation and rising rates hit public and private markets across the board after years of loose monetary policy and a tech bull run.

SoftBank didn’t see that time as a missed opportunity to invest in AI, a person familiar with the company said.

Instead, the the company is of the view that it is still very early in the AI investing cycle, the source added.

Risk and reward

AI technology is fast-moving, from the chips that run the software to the models that underpin popular applications.

Tech giants in the U.S. and China are battling it out to produce ever-advancing AI models with the aim of reaching artificial general intelligence (AGI) — a term with different definitions depending on who you speak to, but one that broadly refers to AI that is smarter than humans. With billions of dollars of investment going into the technology, the risk is high, and the rewards could be even higher.

But disruption can come out of no where.

This year, Chinese firm DeepSeek made waves after releasing a so-called reasoning model that appeared to be developed more cheaply than its U.S. rivals. The fact that a Chinese company managed the feat, despite all the export restrictions for advanced tech in place, rocked global financial markets that were betting the U.S. had an unassailable AI lead.

While markets have since recovered, the potential of surprise advances in technology at such an early stage in AI remains a big risk for the likes of SoftBank.

“As with most technology investments the key challenge is to invest in the winning technologies. Many of the investments SoftBank has made are in the current leaders but AI is still in its relative infancy so other challengers could still rear up from nowhere,” Dan Baker, senior equity analyst at Morningstar, told CNBC.

Still, Son has made it clear he wants to set SoftBank up with DNA that will see it survive and thrive for 300 years, according to the company’s website.

That may go some way to explain the big risks that Son takes, and his conviction when it comes to particular themes and companies — and the valuations he’s willing to pay.

“He (Son) made some mistakes, but directionally he is going in the same driection, which is — he wants to be sure that he is a real player in AI and he is making it happen,” the former Vision Fund exec said.

Continue Reading

Technology

Nvidia and AMD to pay 15% of China chip sales revenues to the U.S. government, FT reports

Published

on

By

Nvidia and AMD to pay 15% of China chip sales revenues to the U.S. government, FT reports

A smartphone with a displayed AMD logo is placed on a computer motherboard in this illustration taken March 6, 2023. 

Florence Lo | Reuters

Nvidia and Advanced Micro Devices have agreed to give the U.S. government a share of revenues from certain chips sold in China, the Financial Times reported, in an unprecedented arrangement with the White House.

In exchange for 15% of revenues from the chip sales, the two chipmakers will receive export licenses to sell Nvidia’s H20 and AMD’s MI308 chips in China, according to the FT.

The arrangement comes as President Donald Trump’s tariffs continue to reverberate through the global economy, underscoring the White House’s willingness to carve out exceptions as a bargaining tool.

Nvidia CEO Jensen Huang met with Trump last week, according to the FT.

In a statement, Nvidia told the Financial Times: “We follow rules the U.S. government sets for our participation in worldwide markets.”

Last week, Trump had said he would implement a 100% tariff on imports of semiconductors and chips, unless a company was “building in the United States.”

Read the complete Financial Times report here.

Continue Reading

Technology

Nvidia claps back against Chinese accusations its H20 chips pose a security risk

Published

on

By

Nvidia claps back against Chinese accusations its H20 chips pose a security risk

Photo illustration of Nvidia’s H20 chip.

Vcg | Visual China Group | Getty Images

Chip giant Nvidia pushed back Sunday in response to allegations from Chinese state media that its H20 artificial intelligence chips are a national security risk for China.

Earlier in the day, Reuters reported Yuyuan Tantian, an account affiliated with Chinese state broadcaster CCTV, said in an article published on WeChat that the Nvidia H20 chips are not technologically advanced or environmentally friendly.

“When a type of chip is neither environmentally friendly, nor advanced, nor safe, as consumers, we certainly have the option not to buy it,” the Yuyuan Tantian article reportedly said, adding that the article said chips could achieve functions including “remote shutdown” through a hardware “backdoor.”

In response, a Nvidia spokesperson told CNBC that “cybersecurity is critically important to us. NVIDIA does not have ‘backdoors’ in our chips that would give anyone a remote way to access or control them.”

Nvidia on Tuesday similarly rejected Chinese accusations that its AI chips include a hardware function that could remotely deactivate the chips, also known as a “kill switch.”

Tensions between the U.S. and China on semiconductor export controls have escalated in recent weeks, even after Nvidia resumed sales of its H20 chip to China. Chinese state media has framed the H20 chip as inferior and dangerous compared to Nvidia’s other chips, while the company has defended its chips.

The company’s resumption of its H20 shipments reversed a previous ban on H20 sales that was placed in April by the Trump administration. Nvidia’s H20 chips — a less-advanced semiconductor compared to its flagship H100 and B100 chips, for example — were developed by Nvidia for the Chinese market after initial export restrictions on advanced AI chips in late 2023.

U.S. export controls on some Nvidia chips are rooted in national security concerns that Beijing could use the more advanced chips to gain an advantage broadly in AI, as well as in its military applications.

Stock Chart IconStock chart icon

hide content

Nvidia stock over the past year.

Chinese officials, meanwhile, are pushing for the U.S. to ease export controls on high-bandwidth memory chips as part of a trade deal before a possible summit between U.S. President Donald Trump and Chinese President Xi Jinping, the Financial Times reported on Sunday, citing people familiar with the matter.

Nvidia CEO Jensen Huang has supported Trump’s policies while also lobbying for export licenses for the H20 AI chip. Huang has said he wants Nvidia to ship more advanced chips to China, underscoring his outspoken stance that Nvidia’s chips becoming the global standard for AI computing is ultimately better for the U.S. to retain market dominance and influence over global AI development.

China is among Nvidia’s largest markets. Nvidia took a $4.5 billion writedown on its unsold H20 inventory in May and has warned that its topline guidance for the July quarter would have been higher by $8 billion without the chip export restrictions.

Nvidia shares were up 1% to close at $182.70 on Friday and are up 36% this year.

Continue Reading

Trending