August 5, 2026 / Case StudiesLegal Knowledge / Read Time: 16 Min

Deleted Company Data to Free Up Space for Freelance Work—and Got Five Years and Ten Months!

A Beijing algorithm engineer deleted roughly 89 TB of AI training data and several text-to-3D models to use company compute for freelance work. He was sentenced to five years and ten months for sabotaging a computer information system. This article examines the conviction, criminal threshold, and calculation of losses.

rm -rf /*

“Drop the database and run” has long been a classic joke among programmers.

This time, however, the person involved neither ran away nor acted out of revenge against his employer.

He deleted the data simply because the company’s servers were full, leaving no room for the freelance work he had taken on.

So he decided to delete the company’s data and free up space for his side project.

In the end, he freed up not only server space, but also the next five years and ten months of his life.

In September 2024, Wang, an algorithm engineer at a technology company in Beijing’s Dongcheng District, wanted to use the company’s computing resources for private freelance work. He improperly logged into the company’s server cluster and executed a high-risk command that forcibly deleted files without confirmation.

Overnight, more than 89 TB of AI training data belonging to the company’s gaming division, together with several internally developed text-to-3D models, was deleted. The model-training system was paralyzed, R&D projects ground to a halt, and the company suffered losses of more than RMB 200,000.

After three court hearings, the court recently convicted Wang of sabotaging a computer information system and sentenced him to five years and ten months in prison.

Why did deleting data worth just over RMB 200,000 result in such a heavy sentence?

Today, let’s look at Beijing’s first criminal case involving the unlawful deletion of AI training data.

*This article reflects only the author’s personal views and is not intended as legal advice.

1. Deleting 89 TB of Company Data for a Freelance Job

Wang was an algorithm engineer at the company—but he was also an insider working against it.

For some time, he had been conspiring with outsiders to use the company’s servers and computing resources to train models for them.

The company’s data cluster, however, was already operating at full capacity for model training and could not run any additional programs.

A normal person would probably have used their own equipment, rented cloud servers, or simply turned down the freelance job.

But when Wang heard that the company planned to migrate its training tasks from Cluster A to Cluster B, he saw an opportunity.

After the migration, Wang checked Cluster A. His own department’s data was gone, but the gaming division’s data remained. He considered downloading a backup, but the server chose that exact moment to—

Freeze.

For reasons that remain unclear, Wang abandoned the backup and entered the “drop-the-database command” instead.

By the time the IT department discovered what was happening, the deletion command had been running for a full seventeen hours.

More than 89 TB of data was deleted outright.

That included all of the gaming division’s R&D materials, AI training data, and several internally developed text-to-3D models.

The company’s model-training system was temporarily paralyzed, bringing the AI gaming division’s R&D projects to a complete halt. After nearly twenty days of work, the gaming division managed to recover most of the deleted data.

The company retained an accounting firm to audit the losses. The audit found that the incident caused more than RMB 160,000 in additional labor costs and over RMB 40,000 in computing-resource expenses related to data recovery.

Wang continued to insist that he had “deleted the data by mistake.” The evidence became conclusive only after the police found messages on his phone showing that he had conspired with outsiders to use the company’s servers for model training, establishing that the deletion was malicious.

The conduct was classified as the crime of sabotaging a computer information system.

2. Why Does Deleting Server Data Constitute Sabotaging a Computer Information System?

Article 286 of China’s Criminal Law expressly addresses the deletion of data from computer systems:

Whoever, in violation of state regulations, deletes, alters, adds to, or interferes with the functions of a computer information system, thereby rendering the system unable to operate normally with serious consequences, shall be sentenced to fixed-term imprisonment of not more than five years or criminal detention. Where the consequences are especially serious, the sentence shall be fixed-term imprisonment of more than five years.

Whoever, in violation of state regulations, deletes, alters, or adds to data or applications stored, processed, or transmitted by a computer information system, with serious consequences, shall be punished in accordance with the preceding paragraph.

Criminal Law of the People’s Republic of China

Here, Wang deleted game-project R&D files, training data, and models stored and processed in the server cluster. His actions also paralyzed the model-training system and halted R&D projects. The conduct plainly falls within the provision above.

Some readers may think that AI training data and models are intangible, were only used to develop games rather than sold externally, and were therefore nothing more than office tools.

“If I delete Microsoft Office, can my company have me arrested?”

But producing a model that can perform real production tasks requires raw materials to be collected, cleaned, deduplicated, and labeled, followed by repeated training and tuning on GPU clusters. The process consumes enormous amounts of labor, time, and computing power.

The resulting model files are no different in substance from a company’s machinery, equipment, raw materials, or other means of production. They simply exist in digital form.

Without those models, the physical servers may still have their lights on, but the original text-to-3D and model-training functions can no longer be performed normally.

Whether a computer information system has been sabotaged therefore cannot be determined solely by asking whether the server was physically smashed or whether the operating system can still boot. The real question is whether the system’s original information-processing functions were materially impaired.

The Supreme People’s Procuratorate later stated that the AI models and training system in this case were treated as a “computer information system.” GPU occupancy time was also used to calculate the computing cost of restoration, providing a new judicial approach to protecting AI models and assessing losses.

3. Why Did Losses of Just Over RMB 200,000 Lead to Five Years and Ten Months?

Some readers may still wonder: he only deleted some files—how did that result in more than five years in prison?

“Was the court somehow siding with the company?”

Not at all.

Under Article 4 of the Interpretation of the Supreme People’s Court and the Supreme People’s Procuratorate on Several Issues Concerning the Application of Law in Criminal Cases Endangering the Security of Computer Information Systems, causing economic losses of at least RMB 10,000 by damaging system functions, data, or applications constitutes “serious consequences.”

Where the amount reaches five times that threshold—RMB 50,000—the consequences are deemed “especially serious,” and the sentence starts at five years.

The losses established in this case exceeded RMB 200,000, well above the threshold for “especially serious consequences.”

The sentence of five years and ten months therefore did not reflect the court “favoring the company.” Wang’s conduct had squarely entered the statutory sentencing range of more than five years.

The calculation of losses also reflects the particular characteristics of the AI industry.

As the saying goes:

How much love can return, and how many people can wait?

After training data is deleted, a company must collect, organize, clean, and label the data again. Restoring the model’s functions also requires renewed investment in GPU resources and engineering hours.

Those are expenses the company would not otherwise have incurred, but was forced to pay because of the deletion. They are real losses in themselves.

The AI industry, moreover, changes every day.

Fall one step behind, and every step that follows may be behind as well.

If a competitor launches a state-of-the-art model first, that may be the “execution line” for your own company.

Of course, “lost business opportunities” may not necessarily form part of the court’s calculation. But the incident occurred in 2024, and at 2026 computing prices, the actual cost of retraining a model alone can already be painful.

4. Final Thoughts

Besides “drop the database and run,” programmers have another classic saying:

“Either the code runs, or the programmer does.”

Unfortunately, people are not very good at running away.

Many may assume that using company resources for a little freelance work would, if discovered, lead at worst to compensation or dismissal.

That is both right and wrong.

It depends on whether the company’s interests have been materially harmed.

Regardless of whether a person has account access, without the company’s written consent, intentionally deleting company data may constitute an infringement of the company’s interests.

But even without deleting the database, merely “stealing compute” may itself constitute a separate criminal offense.

Boyang Li
Author

Boyang Li

Chinese Attorney — Beijing Longan (Guangzhou) Law Firm

A lawyer focused on game law, AI regulation, data compliance, and digital content rights. I write about practical legal insights for innovative tech teams.

Contact me about this topic →