The Looming AI Data Crisis: Turning to Synthetic Data for Solutions

As AI models rapidly consume the internet’s free content, a pressing question emerges: What happens when there’s nothing left to train on?

A recent report from Copyleaks found that DeepSeek, a Chinese AI model, often generates responses nearly identical to ChatGPT. This has raised concerns that it may have been trained on OpenAI’s outputs, highlighting the growing challenge of obtaining high-quality training data.

Some experts suggest that the era of easily accessible, high-value data for AI development may be coming to an end.

In December, Google CEO Sundar Pichai acknowledged this challenge, cautioning that AI developers are quickly depleting the available supply of quality training data.

“In the current generation of LLM models, roughly a few companies have converged at the top, but I think we’re all working on our next versions too,” Pichai said at the New York Times’ Dealbook Summit. “I think the progress is going to get harder.”

The Rise of Synthetic Data

With the availability of high-quality training data shrinking, AI researchers are increasingly turning to synthetic data—artificially generated datasets that mimic real-world information.

Although synthetic data has been used in statistics and machine learning since the late 1960s, its growing role in AI development raises fresh concerns, particularly as AI integrates with decentralized technologies.

“Synthetic data has been around in statistics forever—it’s called bootstrapping,” said Muriel Médard, Professor of Software Engineering at MIT, in an interview with Decrypt at ETH Denver 2025. “You start with actual data and think, ‘I want more but don’t want to pay for it. I’ll make it up based on what I have.”

Médard, co-founder of the decentralized memory infrastructure platform Optimum, emphasized that the primary challenge isn’t data scarcity but accessibility. “You either search for more or fake it with what you have,” she explained. “Accessing data—especially on-chain, where retrieval and updates are crucial—adds another layer of complexity.”

Privacy restrictions and increasing legal protections around real-world datasets are also pushing AI developers toward synthetic data as a viable alternative.

“As privacy restrictions and general content policies are backed with more and more protection, utilizing synthetic data will become a necessity, both out of ease of access and fear of legal recourse,” said Nick Sanchez, Senior Solutions Architect at Druid AI.

“Currently, it’s not a perfect solution, as synthetic data can contain the same biases you would find in real-world data, but its role in handling consent, copyright, and privacy issues will only grow over time.”

Risks and Opportunities

As synthetic data becomes more prevalent, so do concerns about its potential for manipulation and misuse.

“Synthetic data itself might be used to insert false information into the training set, intentionally misleading the AI models,” Sanchez warned. “This is particularly concerning when applying it to sensitive applications like fraud detection, where bad actors could use the synthetic data to train models that overlook certain fraudulent patterns.”

Médard noted that blockchain technology could help mitigate some of these risks by ensuring data integrity. However, she clarified that the goal isn’t to make data unchangeable but rather tamper-proof. “When updating data, you don’t do it willy-nilly—you change a bit and observe,” she said. “When people talk about immutability, they really mean durability, but the full framework matters.”

As AI developers grapple with the diminishing supply of training data, synthetic data is emerging as both a solution and a challenge—offering new opportunities while raising critical ethical and technical concerns.