Our client, a bank operating across the Asia Pacific region, was moving vast volumes of sensitive financial data and AI training datasets into environments they didn’t fully control. This demanded stringent compliance and high levels of security at exactly the point where performance matters most.

The real challenge: these two requirements are usually in tension. Traditional encryption approaches, like IPsec, add packet overhead and can fragment MTU sizes. The problem only increases when the workload is intense AI training or real-time banking transactions that can’t tolerate added latency.

A Senetas CN6140 was installed on-premises to encrypt these very large data volumes at hardware line-rate, adding just single-digit microsecond latency as data left the perimeter. Alongside it, Senetas CV1000 virtual encryptors were spun up as software instances within the bank’s SaaS and AI GPU data lake environments, creating secure encryption endpoints without the need for physical hardware installation.

Together, this architecture eliminated the packet overhead and MTU fragmentation issues common with IPsec, preserving maximum throughput for AI data ingestion pipelines while keeping encryption keys exclusively in the bank’s control.

• Senetas CN6140 hardware encryptor and CV1000 virtual encryptor deployed across on-premises, SaaS and AI GPU environments

• Line-rate, zero-overhead encryption added just single-digit microsecond latency
• Virtual encryption endpoints stood up within SaaS and AI GPU data lakes without physical hardware installation
• Packet overhead and MTU fragmentation issues common with IPsec eliminated
• Maximum throughput preserved for AI data ingestion pipelines
• Encryption keys kept exclusively under the bank’s control throughout