Artificial Intelligence/Machine Learning System Safety

Why AI/ML System Safety?

The technologies and definitions of artificial intelligence and machine learning have come a long way since the 1950s. The one thing that hasn’t changed is the need to assure these systems operate safely within human defined guardrails, obeying hazard controls and mitigations, and assuring human health and safety. These systems must also be designed and operated within the specific operating domain and regulatory framework they are intended.

The Challenge

Artificial Intelligence/Machine Learning is a software and hardware interface system that, for a fixed set of inputs, can produce nondeterministic outputs, approximations, or multiple valid solutions. This makes applying a traditional system safety process complicated, as the boundary of certification is not defined or closely controlled during development and operational phases.

To address AI/ML System Safety specifically, an approach must leverage the best practices in classic hazard analyses techniques, software system safety, data analysis, and understanding the implementation of metaheuristics. Few safety engineers are well versed in all these domains.

Your Solution

APT’s AI/ML System Safety Approach

APT has the breadth of subject matter experts and seasoned engineers from all the necessary system safety disciplines. APT leverages this vast background and experience in solving your AI/ML system safety challenges by applying our innovative AI/ML System Safety Process.

This process reaches into APT’s arsenal of experts in our System Safety Engineering, DOT Competent Authority Approval Services, Test Planning, Software System Safety, Industrial & Occupational Safety & Health, Industrial Engineering, Quality Engineering, Software Development & Modeling, Reliability, Risk Management, and Systems Engineering disciplines.

This is best and most efficiently initiated during the system concept/design phase. However, it is never too late to infuse AI/ML System Safety into your design and operation.

A key to successfully achieving the necessary level of rigor and establishing your safety case lies within APT’s AI/ML System Safety Process implementation side-by-side with your team of system experts.

For more details on APT’s AI/ML System Safety Process, please read our report entitled “A System Safety Process for Artificial Intelligence and Machine Learning Based Products.”

Capabilities

  • System safety program tailoring for AI/ML components
  • Defining AI/ML safety criteria
  • Analyzing AI/ML system hardware/software designs
  • AI/ML training and testing data issue resolution
  • AI/ML risk assessment
  • Plan/implement AI/ML system safety programs
  • AI/ML hazard mitigation and guardrail evaluation
  • AI/ML operational domain evaluation

APT Point of Contact

Wayne Devoid, 321.799.1957
aptinfo@apt-research.com