Softcamp unveils security tools to manage AI agent permissions
The company is launching new solutions to help financial institutions monitor what data AI agents can access and how they interact with external models.
The short version
- SHIELD AI Gateway verifies identities and permissions before allowing AI agents to call large language models.
- SHIELD Gate uses Remote Browser Isolation to separate employee access to external AI from internal business terminals.
- XSCAN Secure Asset uses a Software Bill of Materials to track vulnerabilities in financial service components.
- The solutions are designed to provide clear audit histories of which agent accessed which system and under what authority.
Why it matters
As banks move toward autonomous AI agents, they need a way to ensure these bots don't overstep their boundaries or leak sensitive data. These tools provide the 'guardrails' necessary for high-stakes industries to adopt AI without compromising security.
Securing the AI Agent
Softcamp has announced a new security solution designed specifically for financial institutions that use generative AI. The company plans to showcase these tools at the Financial Information Security Conference, known as FISCON 2026, which takes place on the 12th at the Conrad Seoul. These products focus on managing how AI agents interact with internal systems and external models.
A primary concern for banks is the autonomy of AI agents, which are software programs capable of performing tasks on behalf of humans. Softcamp aims to provide a framework where these agents do not have unrestricted access to sensitive information. By controlling permissions, the company helps firms ensure that AI tools only perform the specific tasks they were assigned to do.
The company's strategy involves creating a clear line of accountability for every action an AI takes. Because AI can process information quickly, it is easy for an agent to move between different databases or tools without human oversight. Softcamp's approach seeks to provide a layer of verification that sits between the AI agent and the data it needs to process.
How SHIELD AI Gateway Works
One of the core products being unveiled is the SHIELD AI Gateway. This system acts as a middleman that manages the process by which AI agents invoke large language models (LLMs) or various business tools. It verifies the identity and specific permissions of an agent before it is allowed to relay a request to a system.
Once the gateway confirms that an agent has the right to perform a specific action, it passes the call through to the destination. Crucially, the gateway also records the entire history of these interactions. This creates a detailed log that shows exactly which AI agent accessed which system and under whose specific authority the request was made.
This level of logging is essential for internal controls and auditing within a bank. If an error occurs or a security breach is detected, administrators can look back at the logs to see the exact path the AI took. It allows human supervisors to maintain a high level of oversight over automated processes that might otherwise be difficult to track.
Preventing Data Leakage
Another tool in the lineup, SHIELD Gate, addresses the risks associated with employees using external generative AI services. It utilizes a technology called Remote Browser Isolation (RBI) to create a barrier between the employee and the business terminal. This separation ensures that when a worker uses an external AI, the data remains isolated from the company's internal network.
The main goal of this technology is to reduce the risk of business data leakage. When employees type sensitive information into a public AI tool, that data could potentially be stored or used by the external provider. By isolating the browser session, SHIELD Gate helps prevent this information from moving from the company's private environment to the public internet.
This type of security is becoming increasingly important as more workers adopt AI to help with daily tasks. Financial institutions are particularly sensitive to this because they handle private customer information and proprietary trade secrets. The isolation method provides a way for employees to stay productive with AI while keeping the company's core data safe.
Managing Software Supply Chains
Softcamp's subsidiary, Redpen Software, is also introducing a tool called XSCAN Secure Asset. This product is based on a Software Bill of Materials, or SBOM. An SBOM is essentially a list of all the components, libraries, and pieces of code that make up a specific software application.
For financial services, knowing every component inside their software is vital for identifying security risks. XSCAN Secure Asset helps these organizations identify the specific components included in their services. If a new vulnerability is discovered in a common software library, the tool can quickly identify which assets are affected by that specific flaw.
The tool also tracks the history of remediation for these vulnerabilities. This means that security teams can see what has been fixed and what still needs attention. By managing the software supply chain in this way, companies can respond more quickly to threats that emerge in the underlying code of their financial platforms.
The Path to Verified AI
Bae Hwan-guk, the CEO of Softcamp, emphasized that financial institutions must be able to verify what data is being sent externally. He noted that companies need to know exactly what tasks AI agents are performing and who authorized those actions. This focus on verification is the central theme of the new security suite being presented at FISCON 2026.
The company is positioning these tools as a necessary step for the wider adoption of AI in the financial sector. Without the ability to audit AI actions and protect data from leaking, many institutions may be hesitant to move beyond simple chatbots. These solutions aim to provide the technical infrastructure needed to make AI a reliable part of the corporate workflow.
Looking ahead, the success of these tools will depend on how well they can keep up with the rapid evolution of generative AI. As models become more complex and agents become more autonomous, the need for robust permission management will only grow. Softcamp is aiming to cover the entire spectrum of these risks, from external access to internal software vulnerabilities.
What to watch
- How other financial institutions adopt these specific permission-based gateways.
- The evolution of SBOM standards for tracking AI-specific software components.
Try it
Check if your company has a policy on using external AI tools and see if they use any isolation technology to prevent data leaks.
Sources and links
- ReportingMoney Today: the article we sourced
First reported by Money Today. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.