Skyflow has launched a new “privacy vault” for large language models (LLMs), which helps developers to improve user data privacy in their applications. The new solution provides enterprises with a layer of security throughout the lifecycle of their LLMs, from data collection to model training and deployment.
Skyflow’s LLM privacy vault uses a proprietary polymorphic encryption technique to ensure sensitive data can be securely handled throughout the LLM process. Companies can define a sensitive information dictionary and have that information protected at all stages of the process. This way, when plaintext sensitive data, such as emails and social security numbers, are provided to GPT models, they will be swapped out with Skyflow-managed tokens and different layers of encryption can be applied. Once the process is completed, the sensitive data is de-tokenized after the GPT model returns its output.
The company is already helping protect sensitive clinical trial data in the drug development cycle and customer data used by travel platforms. IBM is also a user of Skyflow, utilizing the company’s products to de-identify sensitive information.
Skyflow was founded by Anshu Sharma, a former executive at Amazon and Microsoft. Anshu is motivated by the power of bringing the best of technology to bridge the digital divide and bring a secure and equitable digital future for all. The company has received notable investments and is helping keep ahead of the curve when it comes to embedding privacy into applications. On July 11-12, Skyflow will join top executives in San Francisco for a conference on how leaders are optimizing AI investments for success.
