In a significant advancement for artificial intelligence, Ant Group introduced SecretFlow Cloud at the World Artificial Intelligence Conference in Shanghai, a cryptographic computing platform designed specifically for large language models (LLMs). This innovative platform, coupled with a suite of cryptographic solutions, promises to revolutionize data security and privacy protection across cloud-based environments.

SecretFlow Cloud leverages cutting-edge software-hardware cryptographic computing technologies. These technologies enable the secure circulation of encrypted data during both the hosting and inference processes of LLMs, ensuring the protection of valuable LLM assets and enhancing data privacy. With a single click, providers can now encrypt and deploy their LLMs into cloud environments, drastically reducing the risk of data compromise and theft. The encrypted inference service further protects trade secrets during LLM interactions, ensuring efficient data security.

One of the key features of SecretFlow Cloud is its ability to allow GPUs to perform computing tasks within trusted execution environments. This integration significantly narrows the cost and performance gap between encrypted LLM inference and plaintext operations, making it a more cost-effective solution for users. The platform also incorporates memory and disk encryption to provide end-to-end security and facilitates secure cross-domain model hosting management.

SecretFlow Cloud supports both public and private cloud deployments and is compatible with the most commonly used LLMs globally. In a public cloud setup, users can effortlessly build a new professional LLM or migrate an existing one to the platform, with secure LLM inference services ready in as little as 10 minutes. Additionally, the platform offers on-demand cryptographic computing resources, allowing users to scale their operations based on their specific needs.

A user-friendly remote attestation system is another highlight, enabling smooth verification through web interfaces and minimizing user friction. This feature simplifies the process for users to ensure their setups are secure and compliant with data protection regulations.

Dr. Lenx Wei, Vice President and Chief Technology Security Officer at Ant Group, emphasized the importance of data in the application of LLMs. “Cryptographic computing technology not only enhances the utility of data in cross-domain scenarios but is also crucial as LLMs become specialized tools in professional domains,” Dr. Wei noted. He also highlighted that while private environments have traditionally been used to address data security challenges, they often result in higher operational costs and compromised service efficiency. The solutions provided by SecretFlow Cloud can effectively mitigate these issues, unlocking the full potential of LLMs.

Ant Group plans to continue developing its cryptographic computing solutions to ensure comprehensive data security throughout the entire lifecycle of professional LLMs, from creation to deployment and service delivery. Since its initial exploration of privacy-preserving computing technologies in 2016, Ant Group has been at the forefront of this field. The release of SecretFlow in 2022, now followed by SecretFlow Cloud, underscores Ant Group’s commitment to overcoming the pain points associated with security, cost, and accessibility in data circulation. These technologies have already found applications in diverse industries such as insurance, rural finance, healthcare, public services, and marketing, demonstrating their wide-reaching impact.

  AI data security, Ant Group, cryptographic computing, large language models, SecretFlow Cloud