Billionaire Nvidia CEO Jensen Huang has resigned after the tech giant was rocked by reports that he had an affair with a suspected Chinese spy. For days, Jensen Huang has been pilloried after admitting to an affair with a businesswoman who is alleged to be a Chinese spy.
That revelation is bad news for Nvidia, the world’s most valuable company and a major chipmaker integral to the future success of the AI revolution. It was an unfolding scandal that has wiped hundreds of billions of dollars off the chipmaker’s market value, with share prices plunging 10 per cent.
Huang’s problems began when The New York Times reported he had had an affair with former Megaspeed director Alice Huang (no relation), whom US intelligence has said facilitated shipments of Nvidia’s microchips to Chinese firms as part of a foreign intelligence operation. The newspaper subsequently alleged that the executive had also had a relationship with Kimberly Powell, his company’s vice president of healthcare, who is also married.
Huang apologised to his family and colleagues, while Nvidia announced an independent investigation. Subsequently, it claims it has no evidence that neither national security nor corporate secrets were compromised, while China has dismissed the spying allegations as unsubstantiated. Until a permanent successor has been named. Colette Kress, Nvidia’s current executive vice president and chief financial officer, will serve as interim chief executive officer.
It is still unclear whether his wife of 40 years, Lori Huang will stand by him, but beyond the embarrassment of being caught in an affair that affects his wife, family and fellow employees, the incident poses an awkward question for an industry spending billions of dollars developing some of the world’s most valuable technology.
Increasingly it is a victim of sophisticated – and less sophisticated – espionage attacks aimed to try and obtain its inner secrets. So how do you protect against that, and how do you identify that threat when the person being targeted might not even realise what is happening?
In the past, the classic image was a spy trying to penetrate government circles to discover political and military secrets. Today, intelligence agencies also want the knowledge that gives a country its future economic and strategic advantage – and much of that knowledge is held by AI companies, STEM laboratories and universities.
Sometimes the two things collide. In March last year, when Elon Musk was still at the heart of Trump’s administration, Trump said that plans that concerned China should not be shared with Elon Musk because of his business interests.
Trump rejected reports that Musk would be briefed on how the United States would fight a hypothetical war with China, saying: “Elon has businesses in China. And he would be susceptible, perhaps, to that.”
China is central to Musk’s Tesla manufacturing, supply chain, sales and ambitions in autonomous driving and energy storage, and it was a rare admission that the tech billionaire faced conflicts of interest, especially when he was a senior adviser to the US president. While Musk threatened action against the people who had disclosed the proposed briefing, he did not publicly explain how he would manage the underlying conflict Trump had identified.
The issue in this instance wasn’t necessarily that Chinese spies would break into Musk’s systems; it was that a foreign government might have leverage over a businessman whose company depends on its market, factories, suppliers and regulatory goodwill.
For Sam Dunning, director of UK-China Transparency, an organisation that investigates Chinese state influence, the corporate methods that are being deployed are hardly new. “The Chinese government has treated industrial espionage as a major priority for 30-plus years,” he says. “They leverage state resources to conduct espionage on private companies in the West.”
Dunning explains that the methods deployed by China are ones honed over decades of intelligence operations. “It’s the classic espionage tactics like honeytraps, money and vanity,” he says. Tech has simply supercharged the opportunities available to those looking to recruit sources, and China’s push to become a major player in the tech scene means it will do everything necessary to get the upper hand in the tech arms race.
In June, the Five Eyes intelligence alliance, of which the UK is one member alongside the US, Canada, Australia and New Zealand, warned that Chinese intelligence officers were actively posing as recruiters and consultants on networking and job websites like LinkedIn. The targets included individuals who have access to classified or privileged information, who might initially be offered apparently legitimate professional opportunities.
Businesses face two different threats, says David Kilby, a former experienced UK diplomat and executive vice president of corporate espionage firm, Strider Technologies. One is people trying to get a job to steal info from a company, and the other is existing employees being encouraged to turn against their employers after they’ve joined. The former has become increasingly tricky to spot, he says, particularly as AI makes it easier to fabricate convincing identities and employment histories.
For due diligence, Strider investigates potential employees’ backgrounds, poring over publicly available information about their previous work and networks of professional relationships to identify potential issues.
Unlike Huang, who appears to have been targeted at the top of his organisation, useful targets for spies and hostile states can be middle- or lower-ranking officials.
Every so often, we learn about it in the most shocking ways. In October, federal prosecutors charged OpenAI engineer Abdulrahman Alfozan with installing a hidden recording device in an executive conference room at the company’s San Francisco offices. The device was allegedly designed to capture confidential conversations, though Alfozan has yet to be convicted.
Earlier this year, former Google engineer Linwei Ding was convicted of stealing thousands of pages of trade secrets relating to the company’s AI infrastructure while pursuing business interests in China. A jury convicted Ding of economic espionage, but those convictions were later overturned by a federal judge – who found insufficient evidence that he knew his actions would benefit the Chinese government. Ding was sentenced to just under a year in prison for his other crimes. The case showed the risks posed by someone who already has legitimate access to valuable information, and the value to countries like China from those individuals.
That Ding and Alfozan were allegedly engaged in this practice shows how even the stringent procedures big tech firms have in place aren’t always sufficient. OpenAI has said in publicly released documents that once someone is in post, it monitors for unauthorised internal access and suspicious activity, and runs a dedicated team investigating insider threats. Staff device usage and access to databases are tracked to root out any unusual or high-risk activity.
OpenAI competitor Anthropic takes similar approaches; employees there undergo security training and take part in an insider-risk programme to keep things secure. Anthropic says access to valuable model weights – the secret sauce behind its AI models – can require multiple authorisations, hardware security keys and approved devices. It also says it uses controls intended to make the removal of large quantities of information more difficult and easier to detect, but doesn’t detail them for security reasons.
That said, the gap between having a plan and enacting it can be significant. An investigation published last week found OpenAI employees had warned managers about weaknesses in the company’s internal security for years, describing concerns about the accessibility of code and research documentation, and the security of internal tools. OpenAI disputed the claims, saying critical assets were kept in restricted environments and that it had strengthened security measures.
Kilby points to one investigation in which Strider helped a corporate client identify an existing employee who was involved in a Chinese talent programme. According to Kilby, the employee had been using his work laptop to copy and send company information to contacts based in China. The company subsequently examined the employee's laptop to establish what information had been compromised. “It doesn’t happen all the time,” says Kilby. “But it does happen frequently.” Signs of suspicion can even extend to unusual travel patterns or staff logging into corporate systems from unexpected locations. It’s a constant threat.
While staff can be tapped up at the point of applying for jobs within a company, or once they are within a firm, many nation-states choose to play a longer game, trying to get their tendrils into the talent of tomorrow. In September, MI5 warned that the China General Technology Research Institute, an organisation linked to Chinese intelligence, had funded research involving more than 100 UK-linked academics to complete projects in areas like AI and cybersecurity.
Tahlia Peterson, a researcher who has studied foreign interference in Britain’s research sector, says intelligence gathering can be particularly difficult to identify because it may look almost indistinguishable from ordinary academic collaboration. “The bulk of the activity is through innocuous and informal means,” she explains.
Academic involvement can be key because researchers working on cutting-edge technologies often leave academia for better-paid jobs in industry, or develop their research into commercial businesses. “It’s a well-trodden pipeline,” says Peterson. To mitigate against this, the UK government operates a Research Collaboration Advice Team (RCAT), which helps universities assess the national security implications of international research partnerships. RCAT has engaged with more than 155 research institutions and handled over 500 individual cases since its launch in 2022. For its part, Strider identified 5,000 researchers working on STEM subjects connected to Chinese People’s Liberation Army research institutions since 2020.
That’s useful, but Peterson believes support becomes less comprehensive in the world of commercial startups. In her research into protecting British science and technology, this needs a lot more scrutiny and government support.
So how do organisations fix the issue? Peterson reckons screening employees and examining their professional relationships are important parts of the defensive toolkit, but can’t be the only way to spot issues before they arise. There also needs to be better communication between government intelligence agencies, universities and businesses.
It’s also not possible to do everything at once. Instead, Kilby recommends companies think about their “crown jewels” – the information, intellectual property and technologies that would be most damaging to lose – and focus on keeping them safe.
With tech firms increasingly attractive targets for foreign intelligence agencies, Kilby argues: “This is not a discretionary activity – it has to be done,” he says. “But it’s also not a hopeless endeavour.”
