Technical Papers go beneath the policy layer into the engineering and operational detail of making AI safe in occupational settings — written for practitioners who need to implement controls, not just cite standards.
Topics we cover
This section addresses the recurring technical questions that come up across our certification syllabi: how to design meaningful human oversight into an AI-assisted safety system rather than a rubber-stamp "human in the loop"; how to build a failure-mode taxonomy specific to AI (distribution shift, automation complacency, adversarial inputs, silent degradation) rather than reusing generic engineering FMEA categories; and what verification and validation looks like for a model whose behaviour isn't fully specified in advance, unlike traditional safety-critical software.
Explainability as a safety requirement
In safety-critical use, explainability isn't a nice-to-have for user trust — it's what allows an investigator, auditor or supervising professional to determine why a system made a given recommendation or intervention, which is a precondition for effective oversight and for credible post-incident investigation. This is a core competence area for both AISP and AIIP.
Contributing
Members and certified professionals with relevant technical expertise are welcome to propose material for this section — get in touch. In the meantime, related grounding is available in AI Safety Standards and Best Practices.