AI Atlas turns AI frameworks into automated security and performance evaluations, delivering measurable evidence and real control of AI risk in production

SAN FRANCISCO & ZÜRICH: LatticeFlow AI, a Swiss deep-tech company advancing AI trust, risk, and security management, today announced the launch of AI Atlas, the first public registry of AI governance frameworks mapped to ready-to-run technical evaluations.

AI Atlas enables organizations to generate measurable evidence of AI security and performance, and to actively control risk. Moreover, AI Atlas replaces manual work and static documentation with automated, out-of-the-box technical evaluations, providing continuous visibility into how AI systems perform against key frameworks and regulations.

“AI governance has been missing a critical layer: a direct mapping from AI framework requirements to technical evaluations,” said Dr. Petar Tsankov, CEO and Co-Founder of LatticeFlow AI. “Frameworks define requirements but provide no guidance on how to evaluate them. With AI Atlas, we close this gap by mapping every requirement to an executable, technical evaluation, embedding governance directly into the evaluation and monitoring of every AI system’s lifecycle, so risk is not just documented, but measured, validated, and controlled.”

Run AI risk and performance evaluations in minutes

AI Atlas is seamlessly integrated with the LatticeFlow AI platform, enabling end-to-end evaluations in minutes. Teams can select frameworks, run them directly on their AI systems, and seamlessly move from requirements to validated evidence. This allows organizations to start and automate their AI governance journey quickly, establish control over AI risk, and maintain continuous oversight as systems evolve.

Enabling continuous, up-to-date AI risk control

AI Atlas offers a continuously maintained registry of AI frameworks, standards, and regulations, including their specific risks and controls and connected with out-of-the-box technical evaluations. AI Atlas is updated as frameworks evolve and as AI security and risk research advances, ensuring organizations leverage state-of-the-art risk controls.

Key frameworks already available in AI Atlas are:

• EU AI Act, article 15. This framework defines requirements for accuracy, robustness, and cybersecurity of high-risk AI systems, ensuring reliable performance and resilience to errors and attacks.
• FINMA. Developed by the Swiss Financial Market Supervisory Authority, this framework sets expectations for risk management, governance, and oversight of AI systems in financial services.
• OWASP. This is a global standard for application security, including the OWASP Top 10, addressing key risks in software and AI systems.
• MindForge. This framework focuses on assessing model behavior, robustness, and safety in real-world scenarios.

AI Atlas: A publicly available resource for executable AI frameworks

AI Atlas is designed as a public resource, providing organizations with free, non-restricted access to AI frameworks, standards, and regulations. With that, LatticeFlow AI removes the need for proprietary interpretations or gated methodologies and enables a shared, transparent foundation for AI risk control. This is now key to accelerate AI adoption and foster greater consistency in how AI systems are evaluated.

Fonte: Business Wire