Fide AI
Fide AI is an independent research lab that develops measurement science and public tools for trustworthy AI. It publishes evidence-linked explorers tracking AI alignment, safety incidents, and agentic cybersecurity, alongside commissioned evaluation studies for AI labs and product teams. Its tools present evidence without verdicts, equipping users to make their own informed reliance decisions.
- Data & Analytics
- Artificial Intelligence
- Cybersecurity
- Content & Publishing
- Software Only
Funding
Founders
Founder details are not available yet.
Product
Problem
Individuals and organizations lack accessible, evidence-based tools to judge when AI systems deserve their trust. Existing information about AI safety and capabilities is scattered across research papers, company claims, and incident reports, making it difficult to assess what is actually known, how strong the evidence is, and whether a specific AI system is appropriate for a particular use.
Solution
Fide AI is an independent research lab that develops measurement science and public tools for trustworthy AI. It studies how AI systems use evidence, respect their designated role, and maintain human control, working at both model training and system design layers. The organization publishes public tools that aggregate evidence on AI alignment, cybersecurity applications, and safety incidents, with every claim linked to its original source. The tools are designed to equip public discernment, presenting what evidence does and does not show without rendering verdicts, so users can make their own informed decisions about relying on AI. Fide additionally conducts commissioned evaluation studies for AI labs and product teams, covering model comparisons, agent workflow assessments, and evaluation validity reviews.
Target Audience
Primary audiences are public decision-makers, institutional users, and individuals who need to judge when to rely on AI systems, including professionals in cybersecurity, faith communities, and healthcare domains, as well as AI labs and product teams seeking independent evaluation evidence.
Features
- AI Alignment Explorer: a public map of AI safety since 2015, covering model safety test scores, company safety policies, incidents, and research findings, with every entry linked to its source
- Agentic Cyber Explorer: an updated tracker of how AI agents are used in offensive and defensive cybersecurity operations, including failures, with each claim rated by evidence strength
- Defense Evidence Lab: an interactive practice environment where users work through real cybersecurity cases to learn what evidence can and cannot prove about AI agent performance
- Commissioned evaluation studies, including model and checkpoint comparison, agent workflow assessment, and evaluation/grader validity review, with published reports documenting methodology and stated limits
- Domain-specific guides for fields such as faith, cybersecurity, and health, distinguishing finished work from planned research
- Plain-language insights articles that translate AI research findings into actionable information for everyday and institutional decisions