Salesforce Japan Co., Ltd. announced on the 6th “Data Cloud Governance” and “Trusted Services” to support the trust, security, and governance essential for enterprise AI agents. Salesforce explained that to deploy a reliable AI agent, a strong model alone is not enough, but a comprehensive approach based on a secure and well-managed data foundation, safe development practices, and continuous monitoring throughout the AI lifecycle is required. Salesforce’s single architecture integrated platform provides the comprehensive capabilities needed to tackle this challenge head-on. Salesforce’s Backup solution provides automatic backup and accurate restore capabilities to protect 100% of important customer data and metadata stored on Salesforce. Combined with Salesforce Archive, it can meet retention policies while maintaining performance. Realistic data is required to safely develop and test AI agents, but exposing sensitive production data creates risks. Data Mask solutions protect sandbox data at scale, enabling a safe and efficient test environment.
Salesforce Shield enhances visibility into user activity and proactive threat detection capabilities, which are essential for monitoring how AI agents and users interact with data. Security Center simplifies the management of security levels across IT environments with faster data classification and risk assessment. He also explained that beyond core CRM, effective AI agents often need to leverage data across the enterprise, which is why Salesforce Data Cloud is important. Data Cloud brings together all structured and unstructured data to provide a unified, trusted view of the customer that can be leveraged by all front-office applications. Data Cloud Governance enables governance at scale over data and metadata, ensuring consistent policy management and access across Data Cloud. AI-recommended tags allow administrators to automatically label and classify records. For example, data can be marked as “GDPR,” “PII,” or “HIPAA” to ensure data is consistently managed and protected. These tags follow business or compliance frameworks tailored to the needs of any organization. Consistent, granular policies (field, object, record level) can be easily created, managed, and enforced for all data. These policies are automatically applied everywhere in Data Cloud, ensuring data security and consistency across all features, including Agentforce, Analytics, and segmentation. It separates data, metadata and processes by brand, business unit and geography, allowing each business unit to manage its own data while using only one instance of Data Cloud. It also helps create masking policies to maintain the security of sensitive information by automatically hiding or showing data depending on who is accessing it.
In addition, AI agents, including those built with Agentforce, often interact with other systems and data sources through APIs, and ensuring the security of these connections is paramount to prevent vulnerabilities, he explained. With MuleSoft API management solutions, organizations can protect, manage and govern all APIs and enforce security policies and best practices. Importantly for AI agents, MuleSoft Flex Gateway supports protocols such as Agent2Agent (A2A) and Model Context Protocol (MCP) for secure and governed interactions between AI agents and external systems. Confidence in Agentforce deployments is earned through rigorous testing and inherent controls. Salesforce’s highly integrated platform provides the ability to build trust into the Agentforce lifecycle. A secure sandbox environment and Agentforce test center allow for safe development and rigorous testing with realistic data in an isolated environment. Enhancements such as synthetic data generation and state injection speed up this critical stage.
こちらもお読みください： ThinkCyte社、AI細胞解析のためのMorphoScan Cloudを発表
Agentforce itself has features such as variables and filters, and the prompt builder allows agents to control how they reason and what actions they can take based on context. Instruction compliance checks monitor agent behavior against intended instructions and issue alerts if there are deviations , making troubleshooting easier. The Trust Layer will be a core innovation built into the Salesforce Platform. This secure layer sits between the user interface/agent and the large language model (LLM). Features such as automatically removing or masking sensitive data before it leaves the Salesforce boundary for processing by the LLM, preventing sensitive data from being stored by LLM providers, and identifying and flagging potentially harmful or biased language enable reliable and safe use of AI. セールスフォース explains that a platform that can be trusted is needed to move from piloting AI to deploying it across the enterprise. By providing unified data access, powerful developer tools, and robust built-in security and governance, intelligent AI agents and applications, including those built with Agentforce, can be deployed confidently, securely and responsibly to create real business value, improve customer experiences, and increase operational efficiency.
ソース ヤフー
