VECTOR-NATIVE DATA FABRIC
With the rise of AI comes a greater need for vector data environments. This is generating significant interest in autonomous, vector-native data fabrics that can “integrate AI-charged mechanisms directly into the data fabric, authorizing real-time semantic understanding and adaptive data processing,” said Nic Adams, cofounder and CEO at Orcus. Such fabrics are real time and “enable instantaneous retrieval of relevant information without predefined queries.” Such fabrics are destined to lead to a “reorientation in data management,” Adams added. “They embed intelligence directly into their data infrastructures.”
The challenge to implementing such fabrics is in their complexity, Adams cautioned. In addition, he continued, “There is the risk of data poisoning, as a result of malicious inputs. These could corrupt learning models if not properly managed.”
DATA CONTRACTS
With so many enterprise decision makers pinning the future success of their organizations on data, the need for formal data contracts between providers and users of data to ensure its availability and reliability
has become essential. Such contracts can “shift data management from reactive firefighting to proactive governance,” said Kalyan Kumar, chief product officer at HCLSoftware. “The technology creates accountability through automated monitoring and alerting when contract violations occur—whether schema changes, quality degradation, or SLA breaches.” This also includes regulatory compliance requirements.
Data contracts also help to “reduce the time spent on data discovery, cleaning, and reliability issues,” Kumar added. They also help ensure reliable agentic architectures, “providing a clean context for agents to learn how to use data.” According to Kumar, the challenge is “implementing a contract-centric data management practice requires significant cultural transformation, as data contracts demand collaboration between traditionally siloed teams.”
FLEXIBLE ENCRYPTION
Cybersecurity is the topmost priority for data sites—yet, the more data that is encrypted, the more it gets closed off for high-level analytics. An emerging technology called fully homomorphic encryption opens encrypted data to analysis while still maintaining data security and confidentiality.
“Right now, enterprises can only realize a small fraction of their data assets’ value,” said David Archer, CTO at Niobium Microsystems. “For example, McKinsey estimates that organizations could capture 80%–90% more value from financial data if they were empowered to securely share database contents with third parties.”
Enter fully homomorphic encryption, which, “when fully deployed, will enable new open data applications, creating trillions of dollars’ worth of revenue opportunities,” Archer explained. With this technology, “Database contents, queries, and results always remain encrypted. Clients can query databases without revealing their questions, and database owners can share their data without it being exploited. This will clear the way for zero-trust open data applications like dark pool trading, fraud detection, medical research, and more.”
The trade-off, Archer warned, is that fully homomorphic encryption “is both slow and complex.” Currently, fully homomorphic encryption software applications “are still several orders of magnitude slower than computing in the clear. Commercially viable secure data sharing requires hardware acceleration, and the required hardware is only now hitting the market.” This capability may eventually be table stakes for enterprise database environments, he predicted.