Codeficum 42

Measured notes from the workshop and the rack.

How should an AI coding/engineering tool be treated under ISO 26262?

Treat an AI coding or engineering system as a software tool under ISO 26262-8 clause 11. Classify its intended use and the complete workflow around it, not the model name. Tool Impact is TI2 whenever its output enters or influences a safety work product and an error can be introduced or missed. Tool error Detection then depends on checks outside the failure path being assessed; the model does not set its own confidence level.

This is the classification foundation for a three-part series. The companion posts turn it into an operating method for an ISO 26262 project and examine whether AI can provide independent verification evidence.

Same tool, illustrative chain A
AI tool
TI2: it can introduce errors
Independent chain
human review against the specification, independently authored tests, static analysis
Can support TD1
TCL1 if justified
Same tool, illustrative chain B
AI tool
TI2: it can introduce errors
The tool's own loop
its tests, its review, continuous integration as the only gate
Can lead to TD3
TCL3 if justified
Figure 1. Two illustrative outcomes for the same tool. The project assigns TD only after documenting which error classes each external check covers and where correlated gaps remain.

What we reviewed, and on what

This is a desk review of published documents, not a measurement of a system. The document and workstream counts below are checkable against the reference list. The remaining counts are C42 decision templates and gap enumerations, each defined in the corresponding table.

The published set covers the ISO 26262 supporting-process framework and the adjacent AI, governance and autonomous-product sources most directly raised by this question. The workstream set covers four publicly announced revisions or guidance efforts. This is a bounded selection from one review on 2026-08-19, not an exhaustive standards search.

Table 1: what the review covered and what it found, checked 2026-08-19.
Population reviewedCountAddress AI as a development tool
Published standards and technical reports51 (ISO/IEC TR 5469, non-normative)
Standards workstreams in progress41 (ISO/IEC TS 22440, committee draft)
AI tool-use modes we classified64 are TI2 in every case
Places clause 11 gives no guidance12grouped into four families

Laid out in time, the same review makes the gap plainer than the counts do.

202420252026202720282029 review, 2026-08-19 TR 5469 ARP6983 / ED-324 IEC 61508 Ed 3 ISO/PAS 8800 ISO 26262 Ed 3 Q2 to October, tentative TS 22440 CD 2026-04, 2027 likely published or in application expected, tentative a stated window, not a date
Figure 2. The standards review laid out in time. Filled markers are published, hollow markers are expected, and every expected date is tentative. Nothing to the right of the review line is a standard a project can read today.

Scope: ISO 26262-8:2018 clause 11 for tool classification and qualification, ISO 26262-6:2018 clause 9 for software unit verification, the related process standards, and the regulatory landscape as of August 2026.

Disclaimer. This is engineering analysis, not legal advice. Where it discusses product liability or regulatory frameworks, it reports the content of published directives as we understand them. Consult qualified counsel before relying on any statement here for liability or compliance decisions. Consult us for the engineering half: classifying a tool use, building the verification chain that sets TD, writing the qualification plan. The two companion posts show where that classification becomes project work.

Where this text extends past the standard into interpretation, we label it our analysis. No assessment body has validated any interpretation below, and nothing here measures an AI tool: we ran no model and no benchmark for this post.

Apply the decision templates against the licensed edition adopted by the project. The project record should cite the applicable clauses and tables and document any tailoring; the summaries below do not replace that controlled source.

How clause 11 classifies a tool

Clause 11 exists to answer one question: how much do you trust a software tool that touches your safety work products? It does not care whether the tool is a compiler, a static analyser or a large language model. It cares about what happens if the tool is wrong.

Tool Impact (TI) asks whether the tool can introduce or fail to detect an error in a safety work product. TI1 means no. TI2 means yes.

Tool error Detection (TD) asks how likely you are to catch a wrong output before it matters. TD1 is high confidence, TD2 medium, TD3 low or none.

Tool Confidence Level (TCL) combines the two. TI1 tools are always TCL1 whatever TD says. TI2 tools run from TCL1 with TD1 to TCL3 with TD3. Clause 11 is explicit that classification is per use case, not per tool binary, so the same tool used two ways gets two classifications.

One mechanism matters more than any other here, and it is the output verification route. It is not a separate provision. It is what TD means: where activities outside the tool independently verify a TI2 tool's outputs, detection confidence rises toward TD1 and the confidence level falls toward TCL1. Model-based toolchains have run this way for years, which is worth saying before anyone treats this as a special accommodation for AI. The generator is plainly TI2, and the confidence comes from what the project does to the generated code afterwards. The same route is open to an AI tool, on the same condition. The verification has to be genuinely independent and genuinely systematic.

Four qualification methods exist in clause 11: 1a increased confidence from use, 1b evaluation of the tool development process, 1c validation of the software tool, and 1d development in accordance with a safety standard.

1a Confidence from use
Needs version continuity. A model update resets the clock, and a cloud model can update silently.
1b Tool development process
Opaque for a commercial LLM: training data, process and change control are not disclosed.
1c Validation of the tool
Needs defined inputs and outputs. Workable per use case with a fixed suite; temperature 0 does not hold across API versions.
1d Developed to a safety standard
Not for the model. Possible for an in-house orchestration layer.
Figure 3. The four qualification methods against an LLM. For a customer without supplier development evidence, use-case-specific validation under 1c is usually the available route. C42 analysis, checked 2026-08-19.

AI does not stop at code

A model can draft requirements, architecture, detailed design, source code, tests, qualification cases and supporting documents. Clause 11 follows the intended use across all of them: a requirements assistant and a code generator are not different regimes merely because one emits prose.

Interpretation can carry the same impact without producing a file. When a model explains a manual, standard or existing specification and that reading influences a safety decision, the knowledge path reaches the work product. The use is therefore TI2 even if the controlled source is also available; the independent source check contributes to TD instead of changing TI.

The lifecycle map, artefact-specific failures and operational controls move to the practical workflow companion.

Six tool-use decision templates

Artefact type is one axis. Autonomy is the other, and it is the one that moves the detection rating. Table 2 sets out the six modes we see in practice, ordered from the mode where a human checks every output to the mode where automation is the only gate. Read the rows as templates rather than as verdicts: five of the six carry conditional values, because the answer depends on a chain only the project can describe. An LLM used as a suggestion engine and an LLM used as an autonomous agent present different profiles even when they produce the same work product.

Table 2: six AI tool-use decision templates under clause 11. Conditional values require a project-specific argument. C42 analysis, 2026-08-19.
ModeDescriptionTITDTCLRationale
a CopilotHuman reviews every suggestion; full verification downstreamTI2From the named external chainProject-specificThe tool can produce a wrong artefact. Systematic review and independently owned verification can support high detection confidence; their presence does not assign it automatically.
d AI reviewerMarks findings on an artefact that a human also reviewsTI1 if not relied upon, TI2 once influentialFrom the separately accountable verification activityProject-specificAn optional pass can remain TI1 only when neither its findings nor its misses carry the safety argument. Once a decision relies on it, classify that intended use as TI2.
f Retrieval and interpretationReads a manual, standard or specification and reports what it meansTI1 if not relied upon, TI2 once influentialFrom independent source checkingProject-specificThe manual remaining available does not make the AI use TI1. If the reported interpretation influences a safety decision, the tool is TI2; an independent reading of the controlled source contributes to TD.
e AI test generatorProduces test vectors, test code or qualification test casesTI2From independently owned test objectives and oracleProject-specificWrong tests can fail to detect errors. A systematic review against independently owned objectives and expected results can contribute detection evidence; the project still has to justify the rating.
b Agent produces a work productAI drafts requirements, a design or a driver; engineer reviews; the full downstream chain runsTI2From the named downstream chainProject-specificTD follows the systematic checks and the error classes they cover. A written artefact style, source-first review and independently owned verification are evidence inputs, not an automatic rating.
c Autonomous agentProduces, tests, reviews and commits; continuous integration is the only gateTI2TD2 or TD3TCL2 or TCL3If the AI wrote the checks, they are not independent. That correlated failure mode pushes toward TD3.

Two things fall out of that table, and one caveat comes with it. TI is TI2 whenever the output enters or influences a safety work product, and the two TI1 rows hold only while the result is genuinely supplementary and not relied upon: once it influences a safety work product or safety decision, it is TI2. And TD is where the project has agency, because the difference between TCL1 and TCL3 is the chain, not the model. The caveat is that five of the six rows carry conditional values or ranges. They are decision templates for a project to complete against its own chain, not six finished classifications.

Tool output or work product

Clause 11 qualifies tools, not outputs. A compiler is qualified and its object code is verified by testing. An AI generator sits on that boundary, because its output is the same artefact a person would have written, whether that is source code or a requirements specification.

The answer is both. The output is the product of a tool and at the same time a work product entering the project's verification chain. Clause 11 governs the tool. Automotive SPICE governs the work product's lifecycle, and it does not depend on the work product's origin: its process areas specify what must be done to a work product, not who or what produced it. If the process is followed and the evidence trail is complete, an assessment does not fail because a model wrote the first draft.

ToolThe AI generator — classified and qualified under ISO 26262-8 clause 11 as TI, TD, TCLtool regime
OutputThe work product — requirements, architecture, design, code, tests and the documentation that ships with them; lifecycle under Automotive SPICE, verification under ISO 26262 Part 6work-product regime
ChainAI drafts, AI reviews, AI tests, AI documents, automation runs — one composite tool, with TI and TD assessed at pipeline levelclause 11, stretched
Figure 4. Both regimes apply at once: clause 11 to the tool, Automotive SPICE and Part 6 to what it produced. Documentation counts: a safety manual or an integration guide a model drafted is a work product like any other. A chained pipeline is classified as one tool.

When an agent chains tools, clause 11's single-tool framing strains. Our analysis: the most defensible approach today classifies the pipeline as a composite tool, with TI and TD assessed at pipeline level. The current text neither requires nor forecloses that treatment. It is an engineering choice a project makes and records.

Reproducibility, and which kind actually matters

Three kinds of reproducibility get discussed together and they are not equally difficult.

Generation-step reproducibility is not required. If the model writes a different draft each time, but the artefact that ships passes every verification activity the standard mandates, the variation is no different from human engineering. Ask two engineers to implement one specification and you get two implementations. Part 6 verification looks at the output.

Tool qualification reproducibility is the gap. Clause 11 expects tool validation to show that defined inputs produce defined outputs. Method 1a needs evidence of successful use with version continuity, and every model update breaks that continuity. Method 1c needs a suite with defined inputs and expected outputs, and temperature 0 does not guarantee determinism across API versions, quantisation or batching.

Generation step: reproducibility not required
Same input, two runs
two different drafts
Part 6 verification
review, tests, static analysis on what ships
Accepted
as for two engineers
Tool qualification: the real gap
Method 1a
needs version continuity
Method 1c
needs defined outputs
Model update
breaks both
Figure 5. Non-reproducible generation is acceptable when the shipped artefact is verified. Non-reproducible tool behaviour across versions is what clause 11's qualification methods cannot absorb.

The human-engineer analogy has limits. Three asymmetries weaken the argument that a model is just another non-reproducible engineer. A human's competence is assessed under Part 2 and can be audited, while a model's is opaque. Model errors may be correlated across invocations, which undermines the independence assumption. And no validated statistical basis exists for confidence-from-use with a stochastic tool.

Our suggestionDo not rest a qualification argument on method 1a for a tool that changes silently. Run a fixed suite on a schedule instead, built from cases whose expected output a machine can check, and decide in advance what a failure buys you: pin the previous version, or stop using the tool for that use case until it validates again. A suite with no consequence attached is monitoring, not mitigation.

The agentic loop, and where the boundary goes

Put the two halves together and one agent runs the whole thing. It spikes on the target until the device behaves, then drafts requirements from what it learned. From there it proposes an architecture, writes the detailed design, writes the code and writes the tests. It runs them, reads the failures, revises whatever it decides was wrong, drafts the documentation, and hands you the set. Every step is a model making decisions. Where does clause 11 apply? Nothing in the population we reviewed addresses the question, checked 2026-08-19. What follows is our analysis.

PER STEP — every box classified on its own Spike Requirements Design Code Tests Output Qualifies the parts. The decision to loop is not one of them, and neither is the choice of which artefact to revise. PER ITERATION — one pass through the lifecycle Spike Requirements Design Code Tests Output a revision can reach the requirement Each pass looks independent and the errors are not. It also hides the path drawn in red, which leaves every artefact consistent and the product wrong. PER AGENT AS A WHOLE — recommended Spike Requirements Design Code Tests Output One boundary, and it encloses the spike. TI2 for everything handed over, detection from what happens after the agent finishes.
Figure 6. The same pipeline enclosed three ways. The enclosure is the classification boundary, so the three rows are the three answers to where clause 11 applies. Only the last one has a place to put the decision to loop.

Per step is the most faithful to the clause's text, and it fails on the part that matters. Each box is a separate tool use, but the decision to loop is none of them, and neither is the choice of which artefact to revise. You end up qualifying every blade in the blender and not the blender.

Per loop iteration hides the correlated-error problem. Each iteration looks independent and the errors are not. Honest detection rating is TD3, because the detection mechanism inherits the reading that produced the artefact. Across a full-lifecycle workflow it hides something worse: an iteration that resolves a failing test by editing the requirement leaves every artefact consistent and the product wrong.

Per agent as a whole treats the entire agent as one tool: input in, reviewed set out, spike included. Our analysis: this is the only practical framing, because it forces the right questions. TI2, and TD from what happens after the agent finishes. Drawing the boundary here also keeps the spike honest: a spike that stays inside the agent's own loop is not the independent check of the earlier chapter.

Our suggestionClassify the company-standard workflow, not one developer's setup. That means writing the workflow down and putting it in version control first, because a habit cannot be classified and cannot be shown to anyone. Developers will configure agents in their own style whatever you do; the classified unit is the workflow the company defines and imposes.

Agent as tool, not agent as engineer

Agent as tool maps onto clause 11. The agent takes inputs, produces outputs, and a human evaluates them and takes responsibility. This works inside the current standard.

Agent as engineer requires the competence-of-persons requirements in Part 2. The standard speaks of persons with demonstrated competence, training and safety culture. An AI cannot hold competence in that sense, and nothing in the population reviewed supports the framing. Until the standard defines non-human development participants, agent as tool is the only viable reading.

Agent as tool
Inputs
specification, existing artefacts
Agent
Part 8 clause 11: TI, TD, TCL
A named person
evaluates the output and answers for it
Three slots, all filled.
Agent as engineer
Inputs
specification, existing artefacts
Agent
Part 8 clause 11: TI, TD, TCL
The agent again
no legal person answers for it, and Part 2 asks persons for competence, training and safety culture
One substitution, and the third slot empties.
Figure 7. Agent as tool keeps a named person answerable for the output, which is what Part 2 requires. Agent as engineer has no slot for that person, and nothing in the population reviewed supplies one.

A project can classify an agentic system under the current standard with the five steps that follow. Each one leaves a record, and the record is the point: a classification nobody can show is a classification nobody can audit. No assessment body has validated them.

  1. Draw the boundary. Per agent as a whole, input in and reviewed artefact out. Record the boundary definition in version control.
  2. Set TI. TI2 from the harvest line down. Record the rationale against the clause.
  3. Rate TD. From independent checks only, because agent checks are implementation checks. Record the inventory of independent checks, and what each one covers. This step is where the work is.
  4. Read TCL. From the clause 11 matrix. Record the cell you landed in.
  5. Add measures. Approval gate, traceability, drift monitoring. Record which measure answers which residual risk.

Walking the decision tree

The framework collapses into three layers. An engineer gets through them in a single sitting, once per use case, which for one tool on one project usually means more than once.

Can the tool introduce or fail to detect errors in a safety work product?
No — TI1
The project does not rely on the tool's output or its ability to detect errors
TCL1 — document only
Yes — TI2
The tool's output is, or influences, the work product
continue to layer 2
Can downstream verification detect the tool's errors?
High — TD1
Independent review, independently authored checks, qualified analysis, coverage of AI error modes
TCL1 — document only
Medium — TD2
Standard verification, not designed for AI failure modes
TCL2 — qualify the tool
Low or none — TD3
Verification absent, or produced by the same AI
TCL3 — qualify, and evidence the process

C42 policyScale the rigour of the method 1c suite by ASIL: at A and B a representative sample of the project's actual use, at C boundary cases and error injection on top, at D a case in the suite for every AI error mode the project has recorded. This is our own conservative practice, not a reading of the clause, and we state no normative basis for the split. The clause weights its qualification methods by ASIL; these sampling rules are ours.

Layer 3 is the clause 11 matrix, reproduced here in the shape the clause defines rather than as a quotation of it.

Table 3: tool confidence level from tool impact and tool error detection.
TD1TD2TD3
TI1TCL1TCL1TCL1
TI2TCL1TCL2TCL3

Our working reading of the minimum action per level is that TCL1 requires the tool identity, version, configuration, use scope and TI/TD rationale, while TCL2 and TCL3 add one or more qualification methods under the clause's per-ASIL weighting. For a commercial LLM, use-case-specific validation under method 1c is usually the practical route; Figure 3 explains why the other routes are difficult. Verify the final method selection and weighting against the project's licensed copy of the standard.

The standards landscape, compactly

The classification cannot wait for an AI-specific edition of ISO 26262. The published and in-progress documents reviewed on 2026-08-19 divide as follows. Scope statements and previews are not substitutes for unpublished draft text.

Table 4: what each reviewed source contributes to the development-tool question. Desk review, 2026-08-19.
SourceStatus in this argument
ISO 26262-8:2018 clause 11The governing software-tool framework used here; it does not name AI, machine learning or non-deterministic tools.
ISO/PAS 8800:2024Addresses AI as a vehicle component and excludes AI software development tools by scope.
ISO/IEC TR 5469:2024The only published source in the review that classifies AI used during development, through its non-normative B1/B2 usage levels and technology classes.
ISO/IEC 42001:2023 and UL 4600Address AI governance or autonomous-product safety cases, not qualification of an AI development tool under clause 11.
ISO 26262 Edition 3 and IEC 61508 Edition 3In progress. Public previews do not establish an answer for Part 8 tool qualification.
ISO/IEC TS 22440 parts 1 to 3The in-progress workstream whose public scope retains AI used to develop safety-related functions; committee draft as of April 2026.

Regulatory classification of an internal AI development tool is outside this engineering analysis. Obtain that position from counsel and keep it separate from the clause 11 tool-classification argument.

The twelve gaps

Twelve places where ISO 26262-8:2018 clause 11 gives no guidance. Twelve is our enumeration, not a closed set. Numbered for reference, sorted into four families, and not ranked. Table 5 gives, for each one, the working assumption exposed by applying clause 11 and what a large language model does instead.

Table 5: the twelve gaps, the working assumption exposed by applying clause 11 in each and what an LLM does instead, sorted by family. Families 1 and 2 are answerable inside clause 11, family 3 needs Part 9 and Part 2, and family 4 needs a text that does not exist yet. Our assessment, 2026-08-19.
FamilyGapWhat clause 11 assumesWhat an LLM does instead
Version identityG1 Non-deterministic outputDefined inputs produce defined outputsStochastic generation
Version identityG6 Model updatesA tool version can be pinnedCloud-hosted models may change without notice
Version identityG7 Prompt as configurationConfiguration changes are identifiableClause 11 names no prompt change as a re-qualification trigger
Classification inputsG2 Tool chainsA single tool is classifiedAn agentic workflow chains drafting, review, testing and automation
Classification inputsG3 Output or work productThe two regimes are distinctAn AI-produced artefact is both at once
Classification inputsG10 Detection for novel error modesChecks were designed for the error classes presentPlausible-but-wrong output was not in the design basis of those checks
IndependenceG5 CompetenceCompetent persons perform activitiesPart 2 defines no competence equivalent for a model
IndependenceG8 Multi-model pipelinesDistinct tools fail independentlyModels from one lineage share failure modes
IndependenceG9 Independence of AI activitiesVerification is independent of productionOne model produces the artefact and the check on it
IndependenceG11 ASIL decompositionDecomposed elements fail independentlyOne model can put one error pattern into both elements
Outside clause 11G4 Training data and IPTool behaviour has a traceable originTraining data is opaque and may carry licence obligations
Outside clause 11G12 Regulatory scopeNot addressedA counsel-owned question, deliberately outside this engineering classification

G8, G9 and G11 are developed in Can AI provide independent verification evidence under ISO 26262?, which keeps tool independence separate from element independence under decomposition.

Together, these findings establish a narrower conclusion: an AI engineering tool needs no special category under ISO 26262. Classify its intended use and complete workflow. Once its output influences a safety work product, treat the use as TI2 and derive TD from the external verification chain. The second post turns that classification into an operating model; the third examines what can count as independent verification evidence.

References

Clause and part numbers are cited without reproducing the standards' copyrighted text.

  1. ISO 26262-8:2018 — Road vehicles, functional safety, part 8: supporting processes, clause 11 on qualification of software tools
  2. ISO 26262-6:2018 — part 6: product development at the software level, clause 9 on software unit verification
  3. ISO 26262-2:2018 — part 2: management of functional safety, competence management
  4. ISO 26262-9:2018 — part 9: ASIL-oriented and safety-oriented analyses, requirements decomposition and dependent failure analysis
  5. ISO/PAS 8800:2024 — Road vehicles, safety and artificial intelligence
  6. ISO/IEC TR 5469:2024 — Artificial intelligence, functional safety and AI systems, usage levels and technology classes
  7. ISO/IEC CD TS 22440-1, -2 and -3 — functional safety and AI systems: requirements, guidance, examples, committee draft as of 2026-04
  8. ISO/IEC 42001:2023 — AI management system
  9. Automotive SPICE process assessment model 4.0 — SWE.1 to SWE.6 base practices
  10. UL 4600 — Standard for safety for the evaluation of autonomous products
  11. ARP6983, published jointly as ED-324 — aviation guidance for machine learning in airborne systems, in development, target late 2026
  12. IEC 61508 — functional safety of electrical, electronic and programmable electronic safety-related systems