SAS buys synthetic data software from Hazy to streamline AI development – SiliconANGLE

The enterprise data analytics software company SAS Institute Inc. is moving into the realm of synthetic data to boost its artificial intelligence portfolio with the acquisition of intellectual property from the U.K.-based startup Hazy Ltd.

According to SAS, the acquisition will enable it to equip its customers with the tools to create critically needed synthetic data to fuel their AI workloads. The company didn’t provide an acquisition price.

Hazy has created a platform that enables companies to put their most sensitive and private information to use in ways that weren’t previously possible. In many regulated industries, companies have collected enormous volumes of data that could theoretically be used to train highly-performant AI models, but the problem is that they can’t risk exposing that information if they want to stay compliant. Hazy gives them an alternative solution that involves leveraging their existing, highly sensitive data in order to create synthetic information that can be freely used without restrictions.

The synthetic data created using Hazy’s platform mirrors the statistical patterns of real datasets, but it has the advantage of not exposing any private, identifiable or restricted information. As such, it mitigates all of the risk associated with using real data, so companies can feed into their AI models or use it to perform more comprehensive analytics.

SAS said the initial plan is to integrate Hazy’s synthetic data creation tools with the existing SAS Data Maker platform, enhancing its capabilities. It said customers will be able to create much richer synthetic datasets that were previously out of reach, and do so much more rapidly than before, thereby helping to accelerate their AI projects.

With SAS Data Maker, organizations will be able to create more trustworthy AI systems based on diverse synthetic datasets, ensuring they’re extremely reliable. Notably, it provides a viable alternative to the traditional data collection methods, which can be extremely expensive.

SAS Chief Executive Jim Goodnight said the acquisition is a pivotal step in its ambitions to become an industry leader in data management and AI development.

“Hazy is a pioneer in bringing synthetic data to market as a viable enterprise product, and analysts rank it among the top software providers in its category,” Goodnight said. “By integrating their technology, we can offer our customers unparalleled opportunities to harness data safely and effectively, enabling them to experiment and model scenarios that were previously out of reach.”

SAS said the enhanced version of SAS Data Maker will be available in preview in early 2025. Once it becomes available, customers will be able to use their synthetic datasets with its flagship AI platform, SAS Viya.

Announced last year, Viya is an AI integration platform that makes it possible for companies to integrate advanced large language models such as OpenAI’s GPT-4o and Google LLC’s Gemini Pro 1.5 within their existing business processes. In doing this, Viya can then orchestrate those LLMs to create customized AI models and agents and fine-tune them to accomplish very specific business tasks.

International Data Corp. analyst Kathy Lange underscored the importance of synthetic data, saying it can be a game-changer in industries such as healthcare and finance, which are governed by strict privacy regulations. “SAS’ acquisition highlights the growing requirement for synthetic data as an integral component of a modern AI toolkit, addressing data scarcity and privacy issues, and improving model accuracy while reducing biases,” she said.

SAS Chief Technology Officer Bryan Harris discussed the acquisition of Hazy’s software in an interview today on theCUBE with host John Furrier, SiliconANGLE Media’s mobile livestreaming platform. He explained that SAS is making a proactive move, citing studies that show how 75% of businesses are expected to start using generative AI tools to create synthetic customer data by 2026, up from fewer than 5% in 2023.

Image: SiliconANGLE/Freepik AI Suite