Tetrate Expands FINOS AI Governance Framework to Cover Agentic AI Risks, Bolstering Enterprise Defenses Against Agent Hacking

New reference architecture and six new mitigations strengthen protection for AI agents beyond retrieval-based models

SAN FRANCISCO, Oct. 22, 2025 /PRNewswire/ -- Tetrate, a leader in secure microservices and AI infrastructure, today announced a major expansion to the FINOS AI Governance Framework. The move introduces the first comprehensive guidance for agentic AI architectures, a fast-emerging class of AI systems that operate autonomously and interact dynamically with other software and data sources.

The update adds a new reference architecture and six additional risk and mitigation pairs, bringing the total coverage in the FINOS framework to 25. The extension directly addresses new threat vectors such as prompt injection, memory poisoning and persistent agent compromise, which have begun surfacing in enterprise environments. This extended framework is made operational in production environments through Tetrate Agent Operations Director. Serving as a guardrail layer, Agent Operations Director enforces the mitigations defined in the extended FINOS framework and translates governance into operational controls.

For the first time, the two organizations have extended AI governance from simple chatbots to agentic AI. Until now, AI governance frameworks (including FINOS's) were focused on retrieval-augmented generation (RAG) systems with limited autonomy and predictable risk surfaces. But agentic AI systems act on their own, remember past actions and interact with other systems. These capabilities introduce entirely new security and governance challenges.

"This extension underscores FINOS's mission to evolve AI governance in step with AI technologies," said Gabriele Columbro, executive director of FINOS. "By working with contributors like Tetrate, we're assuring that financial institutions and other regulated organizations have a reliable roadmap to stay ahead of the new risks emerging from agentic AI and autonomous systems."

"Agentic AI represents a major architectural shift and with that comes a new set of attack surfaces," said David Wang, head of product management at Tetrate. "Our collaboration with FINOS helps translate the theory of AI governance into concrete, enforceable security practices that organizations can use today."

Bridging the Gap Between AI Ambition and Security Reality
As enterprises adopt more autonomous AI systems, the threat landscape is changing in lockstep. The original FINOS AI Governance Framework was built for retrieval-augmented generation (RAG) models, where risk was relatively contained within predictable workflows. Agentic AI systems, by contrast, operate independently, hold memory over time and can make decisions across multiple contexts, introducing entirely new classes of vulnerabilities.

Tetrate's contribution closes those gaps by defining a vendor-neutral reference architecture and a set of mitigations tailored to the new generation of AI systems. The work focuses on countering issues such as memory poisoning, chain-of-thought leakage and supply-chain compromise, while improving runtime defenses like anomaly detection, behavioral validation and secure memory isolation.

Tetrate also contributed a new visualization tool to the FINOS Common Architecture Language Model (CALM) that delivers full blueprint displays instantly in a modern, user-friendly interface. The tool can be run completely offline, making it easy for financial services users who may have restrictions on connecting to external web services, while requiring minimal setup for easier adoption and updates. Tetrate is the first company outside of Morgan Stanley to contribute code to the CALM project.

These enhancements are designed to give enterprises a standardized, defensible baseline for securing agentic AI deployments before breaches occur. They also make it easier for organizations to align with regulatory and risk management expectations as AI use expands across financial services, government and other regulated sectors.

Modern Threats Demand Modern Governance
Recent industry research, including the MIT 95% Failure Project, has shown that most enterprise AI deployments encounter unexpected failures or adversarial exploits once deployed in real environments. Tetrate's new extension directly addresses this problem by integrating agentic threat models into FINOS's broader AI governance guidance.

The work extends FINOS's coverage from 19 to 25 risks and mitigations, representing one of the most comprehensive public frameworks available for AI governance today. Importantly, the extension was validated using real-world agentic use cases to ensure the guidance is applicable in production, not in theory.

Enabled Through Agent Operations Director
To help organizations implement these new controls, Tetrate developed Agent Operations Director to embed visibility, policy enforcement, anomaly detection and behavioral validation directly into agent runtime environments. By turning policy definitions into automated safeguards, Agent Operations Director gives enterprises a practical way to enforce governance without slowing innovation.

Agent Operations Director marks a significant evolution in how open standards bodies are responding to the realities of AI adoption. For FINOS, it demonstrates a shift toward iterative, community-driven governance, where frameworks evolve continuously alongside technology. For enterprises, it provides a shared reference point to reduce risk, simplify compliance and strengthen defenses against AI-specific attacks.

Together, FINOS and Tetrate are helping organizations move beyond abstract principles and into operational governance—bridging the gap between AI policy, security engineering and production-scale infrastructure.

About Tetrate
Tetrate enables safe, fast and profitable AI transformation by providing trusted connectivity in AI operations. As the driving force behind Envoy and Envoy AI Gateway, Tetrate builds on proven open-source foundations to deliver solutions fit for regulated industries and government organizations. Our products, including Agent Operations Director and Agent Router Service, optimize ROI while establishing trust in AI initiatives. Learn more at www.tetrate.io

About FINOS
FINOS (The Fintech Open Source Foundation) is a nonprofit whose mission is to foster the adoption of open source software, standards and collaborative development practices in financial services. As part of the Linux Foundation, FINOS provides a regulatory-compliant platform for developers from competing organizations to collaborate on innovative projects that transform business operations. With over 100 members spanning major financial institutions, fintechs and technology consultancies, FINOS is at the forefront of driving open source innovation in finance. Get involved and join FINOS as a Member. To stay up to date on FINOS news, events, podcasts, blogs and more, sign up here. Learn more at www.finos.org.

MEDIA CONTACT
Cristin Connelly
Cathey.co for Tetrate
[email protected]

SOURCE Tetrate