ByteDance, the parent company of the popular short video app TikTok, has reportedly fired an intern for sabotaging its artificial intelligence (AI) models. The incident has also exposed the vulnerability of massive data that TikTok holds and has put the spotlight on whether tech giants are taking adequate measures while developing AI.

For context, amid online rumors of ByteDance losing millions of dollars after the intern sabotaged its AI models, the company issued its official statement confirming the incident. It accused the intern of “maliciously interfering” with the training of one of its AI models and indulging in “serious disciplinary violations.”

ByteDance Confirms That It Fired an Intern Over AI Sabotage

Interestingly, ByteDance tried to downplay the impact and said reports of massive damages caused by the actions “contain some exaggerations and inaccuracies.” ByteDance also clarified that “The individual was an intern with the [advertising] technology team and has no experience with the AI Lab.”

It also reacted to the claims made by the unnamed intern and stressed “Their social media profile and some media reports contain inaccuracies.”

Notably, some online reports suggest that over 8,000 H100 GPUs (graphic processing units) were impacted by the sabotage. If these claims are correct, it could translate to a loss of tens of millions of dollars as these Nvidia chips can cost over $25,000 apiece. However, while ByteDance has termed these claims as “exaggerated” it hasn’t specified the monetary loss, and since it is a privately-held company we might never get to know the exact extent of the loss.

ByteDance Says Its LLMs Have Bot Been Impacted by the Sabotage

ByteDance has also refuted the claims made by the intern and said that he did not work at the company’s AI Lab but instead with the commercial technology team. It emphasized that the sabotage had no impact on commercial projects or online businesses as well as its LLMs (large language models).

However, many onlookers don’t buy ByteDance’s side of the story and accuse it of downplaying the incident to hide the real scale of damages. For instance, one commentor suggested that the fired intern used “malicious code” to “deliberately sabotage the training for several months.”

Another commentor on ByteDance’s social media post pointed to the distinction the Chinese tech giant is trying to make between the AI labs and the commercial technology team. They commented, “the commercialization team he is in was previously under the AI Lab. In the past two years, the team’s recruitment was written as AI Lab. He joined the team as an intern in 2021, and it might be the most advanced AI Lab.”

How Could the Sabotage Impact ByteDance?

While we don’t know the exact impact of the sabotage on ByteDance, it could significantly impact the company’s AI strategy. Notably, while the TikTok parent is a leader when it comes to its social media apps and its famously “addictive” algorithm has made its apps quite popular both in the US and China, it initially lagged in AI pivot.

However, its generative AI app Doubao is the most popular app in its category in China, beating out Baidu’s Ernie chatbot. Chinese tech companies are investing aggressively in AI which has led to an “AI war” with China. The US has banned Nvidia from exporting its top-of-the-line AI chips to the communist country which has hit its AI ambitions. Alibaba even scrapped the IPO of its cloud and AI division blaming the US chip export ban.

Work Environment at Chinese Tech Companies Under the Lens

Meanwhile, the ByteDance sabotage incident has put the spotlight on the work environment of Chinese tech companies. Chinese tech companies usually don’t score high on work-life balance and Alibaba’s co-founder is among those Chinese tech leaders who have backed the so-called controversial 996 rule – or working between 9 am-9 pm for six 6 days a week – which implies a 72-hour workweek.

On Glassdoor, while many employees have appreciated the exposure that ByteDance provides, it has been accused of a toxic work environment and extended working hours. Incidentally, a content moderator in the UK sued TikTok earlier this year alleging a toxic work environment at the ByteDance-owned company caused her stress and pregnancy complications.

The bigger question that the ByteDance AI sabotage raises is related to the company’s clearly lacking safety policies.

Data Safety Concerns in AI Development

TikTok is often blamed by Western governments for being a national safety concern because of the vast amount of data that the company possesses. The popular app is battling a ban in the US which it has challenged on First Amendment grounds.

Meanwhile, the sabotage incident has raised an alarm over data safety in AI development. An intern was able to sabotage ByteDance’s AI models which does not speak well for its safety and security standards that the company is observing. No matter how much the Chinese tech giant tries to downplay the financial cost of the sabotage, it has done immense harm to its reputation as a guardian of customer data.

While the sabotage might not have directly impacted TikTok, the incident will provide US authorities more ammunition in targeting the company. The ByteDance AI sabotage might only worsen the regulatory scrutiny that ByteDance is facing in the US where it has an estimated 170 million users.

The sabotage might also lead to calls for better regulatory oversight of AI development so that such incidents don’t occur again.

data breach

Source:

Insider Threat Actors Are a Key Challenge for Companies

The ByteDance sabotage has also highlighted how a threat actor at the right place can potentially do severe damage like stealing the personal data of millions – or even billions of users.

Notably, insider threats, like simple cyberattacks originating from inside the company have been on the rise.

Insiders account for over half of data breaches and with companies processing an ever-increasing quantum of data, the ByteDance sabotage is a wake-up call for other companies also. Would tech companies – which are otherwise pouring billions of dollars into developing AI infrastructure – also get more serious about the safety of data after the ByteDance sabotage? We’ll have to wait and see.