About SafetyCert.AI
Building trust in physical AI as it evolves
Physical AI is changing how machines perceive, decide, and act in the world. The systems being deployed today will not remain static. Their models, software, hardware, applications, and operating environments will continue to evolve.
SafetyCert.AI was founded to ensure that safety assurance can evolve with them.
Why SafetyCert.AI?
Certification must keep pace with innovation
Traditional product certification evaluates a defined product configuration at a particular point in time. That approach works well for products that remain substantially unchanged, but it becomes increasingly difficult to apply to AI-enabled systems that evolve throughout their operating lives.
A change to a model, dataset, sensor, software component, task, or operating environment may affect the assumptions and evidence supporting the original safety decision.
The essential questions are:
What changed?
Which safety claims could be affected?
What evidence remains valid?
What additional testing is required?
Can the product retain its assurance or certification status?
SafetyCert.AI was created to answer these questions through a controlled, independent, and evidence-based process.
Our Purpose
Enable innovation without losing confidence in safety
We help organizations developing physical AI establish credible safety evidence, make informed assurance decisions, and manage change throughout the product lifecycle.
Our work connects four areas that are too often treated separately:
AI and machine learning
Understanding model behavior, data dependencies, performance boundaries, uncertainty, and failure modes.
Functional safety
Connecting AI-enabled behavior to hazards, safety functions, controls, and acceptable residual risk.
Testing and validation
Using explicit acceptance criteria and reproducible experimental decisions to generate defensible technical evidence.
Independent certification
Translating evaluation results into impartial assurance and certification decisions under a defined conformity-assessment framework.
By bringing these disciplines together, SafetyCert.AI helps companies move from promising technical performance to credible evidence of safety.
Our Approach
Dynamic Certification is a lifecycle assurance framework for AI-enabled products that continue to change after their initial evaluation.
It begins by establishing a controlled assurance baseline. When the product changes, the change is evaluated against the safety claims, assumptions, limitations, and evidence supporting that baseline.
Evidence that remains valid is preserved. Claims that may be affected receive targeted review and revalidation. The assurance record is then updated through the appropriate independent decision process.
Establish the baseline. Evaluate the change. Test what is affected. Maintain assurance.
Dynamic Certification is not automatic certification, and it does not mean continuously retesting the entire product. It is a risk-based process for determining what each change requires.
We call it Dynamic Certification
Our Difference
Built specifically for Physical AI
SafetyCert.AI combines AI-specific evaluation methods with the rigor, impartiality, and discipline of established conformity assessment.
Change-aware
We connect product changes to the safety claims, assumptions, and evidence they may affect.
Evidence-preserving
We preserve valid evidence rather than assuming every change requires the complete evaluation to begin again.
Risk-based
We match the level of review and revalidation to the potential safety significance of the change.
AI-specific
We address model behavior, data, uncertainty, robustness, limitations, and performance boundaries.
Reproducible
We use explicit success thresholds to support consistent experimental decisions across repeated trials.
Independent
We provide impartial technical scrutiny that strengthens internal decisions and external confidence.
Our Customers
Supporting the Physical AI ecosystem
We work with organizations developing, integrating, and deploying AI-enabled systems in safety-critical and safety-relevant environments.
Robot and autonomous-system manufacturers
Validate AI-enabled functions and maintain assurance as models, software, tasks, and deployments evolve.
Industrial automation companies
Create a consistent assurance framework across product families, software releases, customer applications, and operating environments.
AI, compute, and safety platforms
Establish repeatable evaluation pathways for products and applications built on a shared technology foundation.
Safety technology providers
Demonstrate how components, subsystems, and software contribute to system-level safety across different integrations.
Operators and technology adopters
Develop credible evidence for deployment decisions involving people, autonomous machines, and changing industrial environments.
Independent Conformity Assessment
From trusted testing to continuing certification
Dynamic Certification requires more than an internal validation process. It requires technical competence, controlled evaluation methods, impartial decisions, and continuing oversight.
SafetyCert.AI is building these capabilities on an internationally recognized conformity-assessment foundation.
ISO/IEC 17025 laboratory accreditation
SafetyCert.AI is completing its ISO/IEC 17025 laboratory accreditation process, with completion targeted for Q4 2026.
The intended capability will support competent and controlled testing of AI-enabled safety functions using defined methods, explicit acceptance criteria, reliable results, and reproducible technical decisions.
Status: Accreditation in progress
Target: Q4 2026
ISO/IEC 17065 product certification accreditation
Following official publication of ISO/IEC TS 22440, SafetyCert.AI plans to complete ISO/IEC 17065 accreditation for its applicable product certification scheme.
The intended capability will support impartial product certification decisions and the continuing maintenance of certification as physical AI products evolve.
Status: Planned following publication of ISO/IEC TS 22440
Target: Q1 2027
Accreditation claims apply only after formal approval and only to activities included within the published scope of accreditation. Target dates remain subject to the applicable standards development and accreditation processes.Our Origin
A new assurance model for a new generation of technology
SafetyCert.AI was founded by Erik Reynolds in response to a fundamental mismatch.
AI-enabled products were advancing rapidly, but the systems used to test and certify them were still largely designed around static product configurations and deterministic behavior.
The industry did not simply need more testing. It needed a way to determine how product change affects safety evidence, which evidence remains valid, and what must be revalidated before the next release or deployment.
That insight became the foundation for Dynamic Certification.
Today, SafetyCert.AI is bringing together expertise in artificial intelligence, functional safety, testing, certification, and international market access to create an independent assurance pathway for physical AI.
Our Leadership
Experience built for the challenge
Erik Reynolds
Founder
As Founder of SafetyCert.AI, Erik Reynolds is advancing a new approach to the assurance of artificial intelligence in safety-critical products. His vision for Dynamic Certification reflects his belief that the methods used to test and certify technology must evolve alongside increasingly adaptive, software-defined, and AI-enabled systems.
Erik has extensive experience in product safety, functional safety, testing, inspection, and certification. Throughout his career, he has worked with manufacturers, technology developers, regulators, and certification organizations to address complex safety and market-access challenges. His experience spans industrial automation, robotics, machinery, electrical products, autonomous systems, and other emerging technologies where innovation is moving faster than established conformity-assessment methods.
Recognizing that conventional certification was designed primarily for static products, Erik founded SafetyCert.AI to develop an independent assurance framework for products that continue to change after their initial evaluation. This work led to Dynamic Certification, a lifecycle approach that connects product changes to affected safety claims, preserves valid evidence, targets revalidation where it is needed, and maintains confidence as AI models, software, hardware, applications, and operating environments evolve.
Throughout his career, Erik has focused on translating emerging technologies into practical safety, testing, and certification programs. He brings together technical experts, industry leaders, research institutions, and conformity-assessment organizations to create pathways that are both scientifically credible and commercially practical.
As a founder and business leader, Erik has built and led organizations serving global technology and manufacturing markets. His approach combines long-term strategic vision with a practical understanding of how safety requirements affect product development, customer acceptance, market access, and the ability to scale.
Erik is motivated by the opportunity to help innovative companies bring transformative technologies to market safely. Through SafetyCert.AI, he is working to make independent assurance a scalable lifecycle capability and establish Dynamic Certification as a foundation for trust in physical AI.
Jo Ihrke
Chief Executive Officer
As Chief Executive Officer of SafetyCert.AI, Jo Ihrke brings a wealth of expertise in product safety, driven by his passion for continuous improvement and innovation.
Holding a Master of Science in Electrical Engineering from the Bundeswehr University Munich, Jo has extensive experience in the Testing, Inspection, and Certification industry, with a particular focus on commercial electrical products. He specializes in supporting business customers through rigorous testing and global certification programs.
His knowledge of Global Market Access requirements is complemented by certification experience across multiple Nationally Recognized Testing Laboratories and leading certification bodies in Germany and Switzerland. Jo’s expertise spans medical, household, industrial, machinery, and information and communications technology products.
Throughout his career, Jo has focused on delivering innovative, customer-centered solutions aligned with market needs. His ability to build and lead highly skilled global teams has consistently supported business growth and operational excellence.
As an experienced profit-and-loss leader, Jo has managed global teams of up to 200 professionals and delivered multimillion-dollar revenue results. His work has included leading greenfield investments in new testing capabilities and rapidly transforming those capabilities into meaningful commercial growth.
Jo is motivated by a deep understanding of market needs and a commitment to helping customers succeed. At SafetyCert.AI, he is building the operational, technical, and conformity-assessment capabilities required to deliver Dynamic Certification at an international scale.
Dr. Patrik Feth
Chief Technical Officer
As Chief Technical Officer of SafetyCert.AI, Dr. Patrik Feth leads the technical development of the company’s Dynamic Certification methodology and its approach to evaluating artificial intelligence in safety-critical systems.
Patrik combines expertise in artificial intelligence, computer science, functional safety, and trustworthy system development. His work focuses on one of the central challenges facing physical AI: translating the probabilistic behavior of AI-enabled systems into explicit safety claims, reproducible evaluation methods, and defensible technical evidence.
In addition to his leadership role at SafetyCert.AI, Patrik is a Professor of Computer Science at IU International University of Applied Sciences. His academic and industry experience enables him to connect emerging research with the practical requirements of product development, conformity assessment, and certification.
At SafetyCert.AI, Patrik is responsible for the technical architecture supporting Dynamic Certification. His work includes developing methods for identifying AI-specific risks, defining measurable acceptance criteria, evaluating robustness and performance boundaries, and determining how changes to models, data, software, hardware, and operating environments affect existing safety evidence.
Patrik is particularly focused on bridging the gap between traditional functional safety and modern AI engineering. His approach recognizes that physical AI requires the discipline of established safety practices while also addressing uncertainty, changing system behavior, software security, and the transition from deterministic to probabilistic engineering.
Working across research, product development, and conformity assessment, Patrik helps transform complex technical concepts into practical testing and certification methods. His leadership ensures that SafetyCert.AI’s services are scientifically rigorous, transparent, reproducible, and scalable across products, releases, and deployment environments.
Patrik is motivated by the opportunity to move trustworthy AI from research into real-world, safety-critical applications. Through SafetyCert.AI, he is helping establish the technical foundation required to evaluate evolving physical AI systems and maintain confidence in their safety throughout the product lifecycle.
Our Principles
Scientific rigor. Commercial practicality. Independent judgment.
Begin with the safety decision
We first identify the decision the evidence must support. This keeps testing focused on meaningful product, deployment, and certification outcomes.
Make success explicit
We define requirements, acceptance criteria, and decision thresholds before evaluation begins.
Preserve traceability
We connect hazards, safety claims, technical requirements, test methods, results, and conclusions.
Report limitations honestly
Credible assurance requires a clear account of where evidence applies, where uncertainty remains, and what conditions could invalidate a conclusion.
Protect impartiality
Our independent judgments are governed by evidence, defined processes, and the applicable conformity-assessment requirements.
Design for the lifecycle
We structure evidence so it can be maintained and extended as products, applications, and operating environments change.
Our Vision
A world where physical AI can evolve safely
Physical AI has the potential to transform mobility, logistics, manufacturing, healthcare, energy, and everyday life. Realizing that potential requires more than increasingly capable machines. It requires a trusted way to determine whether those machines remain safe as they change.
Our vision is to make Dynamic Certification a recognized foundation for trust in physical AI.
We are building toward a future in which safety evidence is current, certification decisions are transparent, valid evidence can be reused, and responsible innovation is not constrained by processes designed for a previous generation of technology.
Work with us
Build your assurance pathway before change becomes a barrier
Whether you are developing one AI-enabled safety function, scaling a family of robots, or creating a platform used across an entire ecosystem, SafetyCert.AI can help you define what must be proven and how that assurance should be maintained.
Begin with a focused conversation about your product, your immediate safety decision, and the changes you expect throughout its lifecycle.