Scan Factory Launches Firewall to Protect Enterprise AI Systems
The product, called Prizma, includes a broad set of tools designed to prevent data leaks.
The rapid adoption of artificial intelligence across businesses clearly brings many benefits. Algorithms speed up numerous processes and help reduce costs. Yet they also introduce risks, particularly a greater potential for data leaks. The market is therefore demanding flexible tools that can protect enterprise AI systems.
A Single Control Point
A notable step in that direction is Prizma, a software platform introduced by Scan Factory. The firewall is designed to protect AI infrastructure. It allows companies to use large language models and agents securely while managing data, access, spending and interactions with external AI services.
Prizma acts as a single control point between users, enterprise systems and AI providers. It analyzes prompts and responses in real time, giving companies centralized oversight of how artificial intelligence is used across their organizations.
Protecting confidential information is one of its core functions. By analyzing user prompts and AI responses, the system automatically flags situations that could lead to data leaks. Risky prompts can also be automatically blocked based on predefined security policies. In addition, the system allows companies to manage AI agents through role-based access and a graph of agent interactions.
Defending Against AI Attacks
The firewall does more than oversee interactions with external services. It also protects against AI-specific attacks. Prizma blocks prompt injection and other attack classes, including threats covered by the OWASP Top 10 for LLM Applications.
The developers have also addressed risks that emerge when companies build enterprise AI infrastructure. In particular, Prizma protects retrieval-augmented generation, or RAG, systems by identifying sensitive corporate information in the files they use. It can also detect shadow AI – unauthorized use of AI services within an organization’s infrastructure. For critical scenarios, a role-based kill switch can quickly restrict individual AI agents or entire chains of interactions.
Building a New Security Layer
“Prizma therefore creates a unified security layer between enterprise infrastructure and AI – from public LLMs to internal models, RAG systems and autonomous agents. With the launch of Prizma, Scan Factory is expanding its cybersecurity business into a new segment for the company – enterprise AI infrastructure protection,” the developer’s press office said.
Enterprise AI security is becoming a new focus for many companies, including those outside Russia. In 2024, Cloudflare announced a firewall for applications that use large language models. It combined tools that could be deployed to monitor and detect vulnerabilities with capabilities already available as part of its WAF, including rate limiting and sensitive-data detection.
The AI Security Toolkit Expands
Russian companies are also actively developing technologies in this area. In 2025, INFERA AI.FireWall, a platform for working with AI services and large language models, entered the market and was officially added to Russia’s domestic software registry. The software protects confidential data, filters malicious and unauthorized prompts, and audits and monitors interactions between users, systems and artificial intelligence.
In February 2026, Solar Group added tools for overseeing the use of large language models to SolarwebProxy. These included policies for employees to interact safely with public and enterprise neural networks, primarily to reduce the risk of data leaks.
Taken together, these developments suggest that a market for enterprise AI security tools is already taking shape in Russia. Experts say the industry needs to develop these technologies further while also helping companies pay greater attention to information security as they adopt artificial intelligence.
