Skip to main content
Open questionOpen question

Open question: how do you evidence a safety case for a controller that was learned rather than specified?

Machinery functional safety asks for a specified function with a quantified failure rate, which a learned policy does not provide. SOTIF supplies a vocabulary for insufficiency, UL 4600 supplies the shape of a safety case, and no published standard tells a robot maker what evidence is enough for a learned controller in a workplace.

The verdict

Open question

No settled answer exists. The ledger poses the question, links the live debate, and does not answer it.

Key facts

What the sources say

Record ID
PAG-2026-0032
Kind
Open question
Jurisdiction
Global
Last verified
Added
  • ISO 13849-1 and IEC 62061 assume a specified safety function with quantifiable reliability in high or continuous demand mode.
  • ISO 21448 addresses functional insufficiencies without failure, and is written for road vehicles.
  • UL 4600 requires a structured safety case for autonomous products and leaves the sufficiency of evidence to argument.
  • The ISO robotics committee's relevant project for dynamically stable mobile robots is still a committee draft.

What it changes

For a robot maker or a deployer

Split the machine: a deterministic, rated safety channel you can certify, and a learned layer that can only propose within limits that channel enforces. Then write the safety case in the UL 4600 shape and the insufficiency analysis in the SOTIF vocabulary. That combination is the strongest evidence available today, and it is an argument, not a certificate.

Sources

What this record was verified against

  1. ISO, ISO 13849-1:2023 Safety related parts of control systemsPrimary · 1 April 2023
  2. ISO, ISO 21448:2022 Safety of the intended functionalityPrimary · 1 June 2022
  3. UL Standards and Engagement, UL 4600 Standard for Evaluation of Autonomous ProductsPrimary · 17 March 2023

Related

Records that sit beside this one

Cite this record

Free to reuse under CC BY 4.0, with attribution. The record ID PAG-2026-0032 is permanent and is never reused.

In a sentence
According to the GAGE Physical AI Governance Atlas (as of 15 September 2026), open question: how do you evidence a safety case for a controller that was learned rather than specified?.
APA
GAGE (Global Academy of Generative-AI Education). (2026). Open question: how do you evidence a safety case for a controller that was learned rather than specified?. Physical AI Governance Atlas. Retrieved 15 September 2026, from https://www.gage.academy/tools/physical-ai-governance-atlas/records/PAG-2026-0032-question-safety-case-for-a-learned-controller
MLA
"Open question: how do you evidence a safety case for a controller that was learned rather than specified?." Physical AI Governance Atlas, GAGE (Global Academy of Generative-AI Education), 15 September 2026, https://www.gage.academy/tools/physical-ai-governance-atlas/records/PAG-2026-0032-question-safety-case-for-a-learned-controller.
Chicago
GAGE (Global Academy of Generative-AI Education). "Open question: how do you evidence a safety case for a controller that was learned rather than specified?." Physical AI Governance Atlas. Last modified 15 September 2026. https://www.gage.academy/tools/physical-ai-governance-atlas/records/PAG-2026-0032-question-safety-case-for-a-learned-controller.
Permalink
https://www.gage.academy/tools/physical-ai-governance-atlas/records/PAG-2026-0032-question-safety-case-for-a-learned-controller

Last updated . Every record re verified . The ledger is checked monthly, first Monday, and the same day for any new law or standard publication.

Back to the full ledger, or every record for Global and every open question record.

GAGE briefings tell you which AI regulation deadlines are coming, what they actually require of you, and when a program opens.

Powered by GAGE, free EU AI Act tools from the Global Academy of Generative-AI Education
CallEmailHelp