BM, in response to the proliferation of agent-based AI, introduces the winning move "Data · AI Trust Layer"

Enterprise Data Protection Company 비엠 Takes the Lead in Building a “Trust Layer” in the AI Era. The company believes that, as generative AI moves toward the widespread adoption of “agent-based AI,” what enterprises truly need is not more powerful computing capabilities, but solutions to the question of “data trustworthiness.”

CEO Anand Eswaran of 비엠 stated at the 비엠온 2026 event that the company is transforming from a backup and recovery enterprise into a “Data and AI Trust Infrastructure” company. He envisions that if NVIDIA handles AI computing, Databricks and Snowflake manage the data layer, and OpenAI and Anthropic develop model intelligence, then 비엠 will build the necessary “trust layer” between them.

Eswaran cited Sam Altman’s remarks to emphasize that the future bottleneck will not be computing or model performance, but whether the “input data to AI is trustworthy.” He explained that above the data layer and below the model layer, an independent data and AI trust layer is needed.

The core of this strategy is the acquisition of Securiti, completed in December 2025. 비엠 acquired Securiti for approximately $1.73B (about 2.5844 trillion Korean won), gaining data security posture management, privacy protection, governance, compliance, and AI trust functions. The company explained that after discussions with enterprise clients, the conclusion was that no matter how sophisticated agent-based AI is introduced, without a robust data management system, achieving expected results would be difficult.

On the same day, 비엠 also announced the “비엠 DataAI Command Platform.” The company claims this is the industry’s first unified data and AI trust infrastructure in the era of agent-based AI. Its core engine, “데이터AI 커맨드 그래프” (DataAI Command Graph), built on Securiti technology, can visualize data at a granular level across cloud, SaaS, and on-premises environments through over 300 connectors.

This means that, unlike existing tools that can only identify at the database or bucket level, it can trace data elements at a finer granularity. 비엠 asserts that this structure not only covers data systems but also serves as a unified knowledge graph encompassing AI systems, identity systems, and agent systems.

The company particularly emphasizes that identity management is inseparable from data security in the age of agent-based AI. The logic is that to truly ensure security and recoverability, one must simultaneously understand which permissions and policies are associated with each data, as well as what content human and non-human accounts have accessed. This allows precise recovery of specific agent misoperations or anomalies occurring within seconds, without rolling back an entire day’s work.

Eswaran calls this “precise resilience.” He explains that when an AI agent makes an incorrect judgment or an automated operation causes an incident, instead of over-rolling back the entire system, selectively restoring only the contaminated behavior can create a significant competitive advantage in enterprise environments.

This trend indicates that the focus of the AI market is shifting from mere model competition to a “operable trust system.” For enterprises, the key to success or failure is less about the introduction of AI itself and more about data quality, governance, access control, and recovery systems. Whether 비엠’s proposed “AI Trust Layer” can become a new standard remains to be seen, but at least during the proliferation of agent-based AI, data trust issues are clearly emerging as a core infrastructure topic.

TP AI Notice: This article is summarized using a language model based on TokenPost.ai. The main content may be incomplete or inconsistent with facts.