Dominion Labs Trust and Safety

Purpose & Scope

Artificial intelligence has the potential to transform industries, elevate human capability, and expand access to knowledge and services worldwide. But like all powerful technologies, AI must be developed with care, foresight, and an unwavering commitment to responsible deployment.

At Dominion Labs, safety is not an afterthought—it is the foundation of every system we build, from lightweight edge models to advanced large-scale reasoning systems. This document outlines our approach to AI safety, responsibility, and ethical stewardship across all products and deployments.

1. Our Commitment to Responsible Innovation

Guided by our Dominion AI Responsibility Principles, we continuously assess, test, and refine our systems against a broad spectrum of safety, security, and ethical risks. Our approach integrates rigorous governance, transparent research practices, and multidisciplinary review to ensure our technologies are aligned with human values and societal well-being.

Our responsibility frameworks are embedded across all of our products—whether TorinAI, Obsidian-32B-Instruct, OE-1, EdgeMED, or specialized domain models—ensuring that every layer of our ecosystem meets the highest standards for reliability, fairness, and safe operation.

2. Governance, Oversight & Internal Review

To safeguard against harm and promote responsible progress, Dominion Labs is establishing an internal oversight body composed of cross-functional leaders in research, security, and ethics.

The oversight body evaluates high-impact research, system updates, external collaborations, and product releases against our Responsibility Principles. This review process ensures that crucial decisions are informed by diverse expertise and that potential risks are identified early and addressed comprehensively.

In parallel, our board evaluates risks associated with advanced model capabilities, including reasoning, autonomy, generalization, and long-horizon decision-making. This board provides guardrails for our most powerful systems and ensures our research trajectory remains aligned with long-term safety considerations.

3. Proactive Safety Measures for Advanced Capabilities

As our models grow more capable, Dominion Labs is investing heavily in proactive safety measures—not just reactive ones. This includes:

• Stress testing models for failure modes across reasoning, perception, and instruction following
• Designing safeguards against misuse in sensitive industries such as healthcare, finance, and public systems
• Developing alignment techniques that ensure model behavior remains predictable, controllable, and grounded in verifiable data
• Building robust monitoring and audit tools for enterprises and developers deploying our models at scale

These efforts reflect our belief that advanced AI must be guided by thoughtful design, empirical rigor, and transparent evaluation.

4. External Collaboration & Public Engagement

We collaborate with researchers and domain experts across technical safety, governance, cybersecurity, and human-computer interaction. These partnerships help deepen the global understanding of AI-related risks and accelerate the development of robust mitigation strategies.

Dominion Labs actively supports open dialogue with policymakers, regulatory bodies, healthcare compliance organizations, and industry leaders to contribute to responsible frameworks for AI deployment. We believe safe AI progress requires collaboration, not isolation.

5. Technical Safety, Security & Frontier Protections

Our research teams focus on:

• Frontier Model Safety – Evaluating emerging capabilities and establishing red lines and safeguards
• Healthcare & Clinical Compliance – Ensuring EdgeMED and EdgeRX meet the highest expectations for HIPAA, PHI, and medical-grade reliability
• Data Security & Encryption – Employing rigorous standards across model training, inference, and storage, including advanced protection for sensitive or regulated data
• Real-World Testing – Validating system behavior across diverse user cases, environments, and stress-conditions

We maintain and continually update the Dominion Frontier Safety Framework—a set of protective measures and best practices designed to ensure our systems remain safe even as their capabilities expand.

6. Long-Term Commitment to Safe AI

Our interdisciplinary teams across research, engineering, clinical science, ethics, and public engagement are committed to ensuring that AI technology advances responsibly, inclusively, and safely.

We are investing in foundational research, large-scale evaluations, multi-level governance systems, and proactive safety tooling to help shape a future where AI empowers humanity without compromising security, fairness, or trust.

At Dominion Labs, we view responsible AI not as a requirement but as a shared obligation to society, and a core part of our identity as a frontier research organization.