feat(ai-sdk): classify technical standards (NIST/OWASP/Grundschutz) as technical_standard

The existing NIST corpus (SP 800-82r3 etc., ingested before source_class tagging)
was classified supervisory_guidance because "NIST" sat in guidanceMarkers, so the
control-intent lift (#36) could never surface it. Add a technical_standard class:

- authority.go: new standardMarkers (NIST/OWASP/Grundschutz/ISO 27001/CSA CCM/CIS),
  checked before guidanceMarkers (so "BSI Grundschutz" -> standard, not BSI guidance);
  move NIST out of guidanceMarkers; sourceClassFromWeight maps weight 80 -> standard.
- authority_rerank.go: the intent-lift path (liftAboveBinding + bestBindingSemantic)
  now classifies via classifyAuthority instead of trusting the raw payload source_class,
  so the untagged legacy corpus is recognized — untagged NIST is now lifted on a
  control question ("Welche Controls passen zu Security Updates?").

Tested: classifier cases for NIST/Grundschutz/weight-80, and an untagged-NIST lift case.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

ai-compliance-sdk/internal/ucca/authority.go

@@ -9,8 +9,8 @@ import (
 // authorityInfo is the normative classification of a search result, used internally
 // for re-ranking only (Phase 1 changes ordering, not the response contract).
 type authorityInfo struct {
-	weight       int    // 100 binding_law, 70 guidance, 0 foreign_law, 50 unknown
-	sourceClass  string // binding_law | supervisory_guidance | foreign_law | unknown
+	weight       int    // 100 binding, 80 technical_standard, 70 guidance, 0 foreign, 50 unknown
+	sourceClass  string // binding_law | technical_standard | supervisory_guidance | foreign_law | unknown
 	jurisdiction string // DE | EU | CH
 }
 
@@ -18,7 +18,13 @@ var (
 	guidanceMarkers = []string{
 		"DSK", "EDPB", "BfDI", "BFDI", "BayLfD", "Baylfb", "ENISA", "BSI", "EUCC",
 		"Standards Mapping", "Kpnr", "Orientierungshilfe", "Handreichung", "Beschluss",
-		"Leitlinie", "Guidance", "Empfehlung", "NIST", "OECD", "CISA", "Blue Guide",
+		"Leitlinie", "Guidance", "Empfehlung", "OECD", "CISA", "Blue Guide",
+	}
+	// Technical standards / control frameworks (best-practice controls). Checked BEFORE
+	// guidanceMarkers so a "BSI Grundschutz" chunk classifies as a standard, not BSI guidance.
+	standardMarkers = []string{
+		"NIST", "OWASP", "Grundschutz", "ISO 27001", "ISO/IEC 27001",
+		"CSA CCM", "Cloud Controls Matrix", "CIS Benchmark", "CIS Control",
 	}
 	foreignMarkers = []string{"RevDSG", "fedlex", "(CH)"}
 	deMarkers      = []string{"BDSG", "DSK", "BfDI", "BFDI", "BayLfD", "Baylfb", "BSI"}
@@ -48,6 +54,8 @@ func classifyAuthority(r LegalSearchResult) authorityInfo {
 	switch {
 	case containsAny(hay, foreignMarkers):
 		return authorityInfo{weight: 0, sourceClass: "foreign_law", jurisdiction: "CH"}
+	case r.Category == "standard" || containsAny(hay, standardMarkers):
+		return authorityInfo{weight: 80, sourceClass: "technical_standard", jurisdiction: jur}
 	case r.Category == "guidance" || containsAny(hay, guidanceMarkers):
 		return authorityInfo{weight: 70, sourceClass: "supervisory_guidance", jurisdiction: jur}
 	case r.Category == "regulation" || r.Category == "eu_recht" || normPattern.MatchString(r.ArticleLabel):
@@ -61,6 +69,8 @@ func sourceClassFromWeight(w int) string {
 	switch {
 	case w >= 100:
 		return "binding_law"
+	case w >= 80:
+		return "technical_standard"
 	case w >= 70:
 		return "supervisory_guidance"
 	case w <= 0:

ai-compliance-sdk/internal/ucca/authority_rerank.go

@@ -7,17 +7,17 @@ import (
 
 // Re-ranking coefficients (validated in the offline golden harness; Phase A — conservative).
 const (
-	authorityCoef        = 0.40 // * weight/100
-	jurisdictionGain     = 0.05 // binding/guidance from DE or EU
-	foreignPenalty       = 0.60 // foreign law on a DE/EU question (demoted, not removed)
-	unknownPenalty       = 0.08
-	domainMatchGain      = 0.15
-	offDomainPenalty     = 0.10 // off-domain binding (demoted, not removed)
-	scopePenalty         = 0.25 // BDSG Teil 3 (law enforcement) on a general DP question
-	topicGain            = 0.18 // amplifier only
-	supersededPenalty    = 0.50 // superseded Alt-Quelle (pre-eu-v1): demoted, nicht versteckt
-	guidanceIntentGain   = 0.10 // epsilon a qualifying guideline is lifted ABOVE the best binding hit
-	guidanceIntentMargin = 0.05 // ...only if the guideline is semantically competitive with binding
+	authorityCoef     = 0.40 // * weight/100
+	jurisdictionGain  = 0.05 // binding/guidance from DE or EU
+	foreignPenalty    = 0.60 // foreign law on a DE/EU question (demoted, not removed)
+	unknownPenalty    = 0.08
+	domainMatchGain   = 0.15
+	offDomainPenalty  = 0.10 // off-domain binding (demoted, not removed)
+	scopePenalty      = 0.25 // BDSG Teil 3 (law enforcement) on a general DP question
+	topicGain         = 0.18 // amplifier only
+	supersededPenalty = 0.50 // superseded Alt-Quelle (pre-eu-v1): demoted, nicht versteckt
+	intentLiftGain    = 0.10 // epsilon a qualifying interpretative source is lifted ABOVE the best binding
+	intentLiftMargin  = 0.05 // ...only if that source is semantically competitive with binding
 )
 
 // guidanceIntentSignals mark a query that EXPLICITLY asks for an interpretation /
@@ -29,10 +29,19 @@ var guidanceIntentSignals = []string{
 	"auslegung", "empfiehlt", "empfehlung", "sagt", "laut",
 }
 
-// queryWantsGuidance reports whether the query explicitly asks for guidance/interpretation.
-func queryWantsGuidance(query string) bool {
+// controlIntentSignals mark a query that asks HOW to implement / which controls or
+// measures fit — rather than WHAT the binding obligation is. Only then may a
+// (semantically competitive) technical_standard outrank the binding norm.
+var controlIntentSignals = []string{
+	"control", "controls", "maßnahme", "massnahme", "schutzmaßnahme",
+	"best practice", "best-practice", "umsetzen", "implementier", "absicher",
+	"härt", "haert", "hardening", "nist", "owasp", "grundschutz",
+	"ccm", "iso 27001", "isms",
+}
+
+func queryMatchesAny(query string, signals []string) bool {
 	q := strings.ToLower(query)
-	for _, sig := range guidanceIntentSignals {
+	for _, sig := range signals {
 		if strings.Contains(q, sig) {
 			return true
 		}
@@ -40,16 +49,22 @@ func queryWantsGuidance(query string) bool {
 	return false
 }
 
+// queryWantsGuidance reports whether the query explicitly asks for guidance/interpretation.
+func queryWantsGuidance(query string) bool { return queryMatchesAny(query, guidanceIntentSignals) }
+
+// queryWantsControls reports whether the query asks for implementation controls/measures.
+func queryWantsControls(query string) bool { return queryMatchesAny(query, controlIntentSignals) }
+
 // bestBindingSemantic returns the highest RAW semantic score among binding-law
-// results (0 if none / intent not requested). Used as the guard threshold so an
-// off-topic guideline cannot ride the interpretation-intent boost.
-func bestBindingSemantic(results []LegalSearchResult, wantsGuidance bool) float64 {
-	if !wantsGuidance {
+// results (0 if none / no intent). Used as the guard threshold so an off-topic
+// interpretative source cannot ride the intent boost.
+func bestBindingSemantic(results []LegalSearchResult, wantsIntent bool) float64 {
+	if !wantsIntent {
 		return 0
 	}
 	best := 0.0
 	for _, r := range results {
-		if r.SourceClass == "binding_law" && r.Score > best {
+		if classifyAuthority(r).sourceClass == "binding_law" && r.Score > best {
 			best = r.Score
 		}
 	}
@@ -104,15 +119,22 @@ func rerankByAuthority(query string, results []LegalSearchResult) []LegalSearchR
 	qDomain := queryDomain(query)
 	qForeign := queryIsForeign(query)
 	wantsGuidance := queryWantsGuidance(query)
-	bestBindingSem := bestBindingSemantic(results, wantsGuidance)
+	wantsControls := queryWantsControls(query)
+	bestBindingSem := bestBindingSemantic(results, wantsGuidance || wantsControls)
 
 	out := make([]LegalSearchResult, len(results))
 	copy(out, results)
 	for i := range out {
 		out[i].Score = authorityScore(query, out[i], qDomain, qForeign)
 	}
+	// Explicit interpretation intent → a competitive guideline may outrank binding;
+	// explicit implementation intent → a competitive technical_standard may. Both lift
+	// ABOVE the best binding FINAL, so a pure norm question (neither intent) is untouched.
 	if wantsGuidance {
-		applyGuidanceIntent(out, results, bestBindingSem)
+		liftAboveBinding(out, results, bestBindingSem, "supervisory_guidance")
+	}
+	if wantsControls {
+		liftAboveBinding(out, results, bestBindingSem, "technical_standard")
 	}
 	sort.SliceStable(out, func(a, b int) bool {
 		return out[a].Score > out[b].Score
@@ -120,24 +142,27 @@ func rerankByAuthority(query string, results []LegalSearchResult) []LegalSearchR
 	return out
 }
 
-// applyGuidanceIntent lifts semantically-competitive guidance just ABOVE the best
-// binding hit (ordered by semantic), so an EXPLICIT interpretation question can
-// return guidance Top-1. Obligation questions (no intent → not called) keep
-// binding on top. Guidance below the semantic margin is left untouched, so an
-// off-topic guideline can never ride the override — and the lift is computed from
-// the binding FINAL score, so authority/topic/domain bonuses cannot edge it out.
-func applyGuidanceIntent(out, raw []LegalSearchResult, bestBindingSem float64) {
+// liftAboveBinding lifts a semantically-competitive interpretative source (the given
+// sourceClass — supervisory_guidance or technical_standard) just ABOVE the best binding
+// hit, ordered by semantic, so an EXPLICIT guidance/implementation question can return
+// that source Top-1. A pure norm question (no intent → not called) keeps binding on top.
+// Sources below the semantic margin are left untouched, so an off-topic source can never
+// ride the override — and the lift is from the binding FINAL score, so authority/topic/
+// domain bonuses cannot edge it out.
+func liftAboveBinding(out, raw []LegalSearchResult, bestBindingSem float64, sourceClass string) {
 	bestBindingFinal := 0.0
 	for i := range out {
-		if out[i].SourceClass == "binding_law" && out[i].Score > bestBindingFinal {
+		if classifyAuthority(out[i]).sourceClass == "binding_law" && out[i].Score > bestBindingFinal {
 			bestBindingFinal = out[i].Score
 		}
 	}
 	for i := range out {
-		if out[i].SourceClass != "supervisory_guidance" || raw[i].Score < bestBindingSem-guidanceIntentMargin {
+		// Classify (not raw payload) so the untagged legacy corpus — e.g. NIST ingested
+		// before source_class tagging — is still recognized as its interpretative class.
+		if classifyAuthority(out[i]).sourceClass != sourceClass || raw[i].Score < bestBindingSem-intentLiftMargin {
 			continue
 		}
-		lifted := bestBindingFinal + guidanceIntentGain + (raw[i].Score - bestBindingSem)
+		lifted := bestBindingFinal + intentLiftGain + (raw[i].Score - bestBindingSem)
 		if lifted > out[i].Score {
 			out[i].Score = lifted
 		}

ai-compliance-sdk/internal/ucca/authority_test.go

@@ -14,6 +14,10 @@ func TestClassifyAuthority(t *testing.T) {
 		{"tagged guidance DE", LegalSearchResult{AuthorityWeight: 70, SourceClass: "supervisory_guidance", Jurisdiction: "DE"}, 70, "supervisory_guidance", "DE"},
 		{"tagged foreign CH", LegalSearchResult{AuthorityWeight: 0, SourceClass: "foreign_law", Jurisdiction: "CH"}, 0, "foreign_law", "CH"},
 		{"untagged ENISA guidance", LegalSearchResult{RegulationShort: "ENISA", ArticleLabel: "ENISA CRA Standards Mapping"}, 70, "supervisory_guidance", "EU"},
+		{"untagged NIST standard", LegalSearchResult{RegulationShort: "NIST SP 800-82r3", ArticleLabel: "AU-8"}, 80, "technical_standard", "EU"},
+		{"BSI Grundschutz standard beats BSI guidance", LegalSearchResult{RegulationShort: "BSI Grundschutz", ArticleLabel: "BSI Grundschutz Baustein"}, 80, "technical_standard", "DE"},
+		{"weight-only 85 TRGS standard", LegalSearchResult{AuthorityWeight: 85, RegulationShort: "TRGS 529"}, 85, "technical_standard", "EU"},
+		{"tagged technical_standard", LegalSearchResult{AuthorityWeight: 80, SourceClass: "technical_standard", Jurisdiction: "EU"}, 80, "technical_standard", "EU"},
 		{"untagged CRA binding", LegalSearchResult{RegulationShort: "CRA", ArticleLabel: "Art. 13 CRA", Category: "regulation"}, 100, "binding_law", "EU"},
 		{"untagged BDSG binding DE", LegalSearchResult{RegulationShort: "BDSG", ArticleLabel: "§ 38 BDSG"}, 100, "binding_law", "DE"},
 		{"untagged RevDSG foreign", LegalSearchResult{RegulationShort: "RevDSG", ArticleLabel: "RevDSG (CH)"}, 0, "foreign_law", "CH"},

ai-compliance-sdk/internal/ucca/legal_rag_intent_test.go

@@ -70,3 +70,79 @@ func TestRerank_OffTopicGuidance_BlockedByGuard(t *testing.T) {
 		t.Errorf("off-topic guidance must not win even with intent, got %s", out[0].SourceClass)
 	}
 }
+
+func TestQueryWantsControls(t *testing.T) {
+	wants := []string{
+		"Welche Controls passen zu Security Updates?",
+		"Welche Maßnahmen sollten wir umsetzen?",
+		"Wie härten wir den Server ab?",
+		"Gibt es NIST-Controls dafür?",
+		"OWASP Best Practice für Logging?",
+		"BSI Grundschutz Bausteine",
+	}
+	plain := []string{
+		"Welche Anforderungen bestehen an Security Updates?",
+		"Ab wann braucht man einen Datenschutzbeauftragten?",
+	}
+	for _, q := range wants {
+		if !queryWantsControls(q) {
+			t.Errorf("should detect control/implementation intent: %q", q)
+		}
+	}
+	for _, q := range plain {
+		if queryWantsControls(q) {
+			t.Errorf("should NOT detect control intent (norm question): %q", q)
+		}
+	}
+}
+
+func TestRerank_ControlQuestion_StandardMayWin(t *testing.T) {
+	// Explicit implementation intent + standard semantically competitive → standard wins.
+	results := []LegalSearchResult{
+		intentRes("NIST SP 800-82", "technical_standard", 0.62, 80),
+		intentRes("CRA", "binding_law", 0.58, 100),
+	}
+	out := rerankByAuthority("Welche Controls passen zu Security Updates?", results)
+	if out[0].SourceClass != "technical_standard" {
+		t.Errorf("control question: technical_standard should win Top-1, got %s", out[0].SourceClass)
+	}
+}
+
+func TestRerank_NormQuestion_BindingOverStandard(t *testing.T) {
+	// "Anforderungen" → no control intent → binding stays Top-1 over the standard.
+	results := []LegalSearchResult{
+		intentRes("NIST SP 800-82", "technical_standard", 0.62, 80),
+		intentRes("CRA", "binding_law", 0.58, 100),
+	}
+	out := rerankByAuthority("Welche Anforderungen bestehen an Security Updates?", results)
+	if out[0].SourceClass != "binding_law" {
+		t.Errorf("norm question: binding must stay Top-1 over standard, got %s", out[0].SourceClass)
+	}
+}
+
+func TestRerank_OffTopicStandard_BlockedByGuard(t *testing.T) {
+	// Control intent present, but the standard is semantically far below binding →
+	// the margin guard keeps binding Top-1 (no off-topic standard override).
+	results := []LegalSearchResult{
+		intentRes("NIST SP 800-82", "technical_standard", 0.40, 80),
+		intentRes("CRA", "binding_law", 0.58, 100),
+	}
+	out := rerankByAuthority("Welche Controls passen zu Security Updates?", results)
+	if out[0].SourceClass != "binding_law" {
+		t.Errorf("off-topic standard must not win even with control intent, got %s", out[0].SourceClass)
+	}
+}
+
+func TestRerank_ControlQuestion_UntaggedNISTLifted(t *testing.T) {
+	// The existing NIST corpus is UNtagged (no source_class). It must still be classified
+	// technical_standard via markers and lifted on a control question — the whole reason
+	// the lift path classifies instead of trusting the raw payload field.
+	results := []LegalSearchResult{
+		{RegulationShort: "NIST SP 800-82r3", ArticleLabel: "AU-8", Score: 0.62},
+		{RegulationShort: "CRA", ArticleLabel: "Art. 13 CRA", Category: "regulation", Score: 0.58},
+	}
+	out := rerankByAuthority("Welche Controls passen zu Security Updates?", results)
+	if out[0].RegulationShort != "NIST SP 800-82r3" {
+		t.Errorf("untagged NIST should be lifted Top-1 on a control question, got %q", out[0].RegulationShort)
+	}
+}