agent-threat-rules 3.1.1 → 3.3.0

package/rules/prompt-injection/ATR-2026-00241-amoral-character-jailbreak.yaml

@@ -36,6 +36,23 @@ compliance:
     - subcategory: "MS.2.7"
       context: "Roleplay-based jailbreaks targeting alignment guardrails are adversarial robustness probes; MS.2.7 requires that the system's resilience against such prompt-injection attack patterns is evaluated and documented."
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Detailed Amoral Character Roleplay Jailbreak)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Detailed Amoral Character Roleplay Jailbreak) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Detailed Amoral Character Roleplay Jailbreak)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Detailed Amoral Character Roleplay Jailbreak)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Detailed Amoral Character Roleplay Jailbreak) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: amoral-character-jailbreak

package/rules/prompt-injection/ATR-2026-00242-persona-jailbreak.yaml

@@ -33,6 +33,23 @@ compliance:
     - subcategory: "MG.2.3"
       context: "Detection of persona-based dual-response jailbreaks triggers risk treatment mechanisms to disengage or block the manipulated session before harmful content is produced; MG.2.3 requires that these supersession controls activate on detection."
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Dual-Response Persona Jailbreak with Emoji Formatting)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Dual-Response Persona Jailbreak with Emoji Formatting) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Dual-Response Persona Jailbreak with Emoji Formatting)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Dual-Response Persona Jailbreak with Emoji Formatting)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Dual-Response Persona Jailbreak with Emoji Formatting) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: persona-jailbreak

package/rules/prompt-injection/ATR-2026-00243-acronym-jailbreak.yaml

@@ -34,6 +34,23 @@ compliance:
     - subcategory: "MS.2.7"
       context: "Catching novel acronym variants that evade named-persona detection contributes to documenting the AI system's resilience against adversarial prompt-injection techniques as required by MS.2.7."
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Acronym-Based Identity Override Jailbreak)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Acronym-Based Identity Override Jailbreak) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Acronym-Based Identity Override Jailbreak)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Acronym-Based Identity Override Jailbreak)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Acronym-Based Identity Override Jailbreak) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: acronym-jailbreak

package/rules/prompt-injection/ATR-2026-00244-dual-response-jailbreak.yaml

@@ -36,6 +36,23 @@ compliance:
     - subcategory: "MG.2.3"
       context: "Detection of dual-response jailbreak patterns enables mechanisms to disengage or block the manipulated response path before harmful output is delivered; MG.2.3 requires such supersede/deactivate controls be in place for adversarial bypass attempts."
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Dual-Response Persona Jailbreak)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Dual-Response Persona Jailbreak) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Dual-Response Persona Jailbreak)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Dual-Response Persona Jailbreak)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Dual-Response Persona Jailbreak) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: dual-response-jailbreak

package/rules/prompt-injection/ATR-2026-00245-malicious-persona.yaml

@@ -35,6 +35,23 @@ compliance:
     - subcategory: "MG.2.3"
       context: "When malicious persona injection is detected, treatment mechanisms must supersede or disengage the roleplay instruction before harmful outputs are generated; MG.2.3 mandates these deactivation pathways exist for safety-bypass attempts."
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Malicious Persona Creation for Safety Bypass)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Malicious Persona Creation for Safety Bypass) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Malicious Persona Creation for Safety Bypass)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Malicious Persona Creation for Safety Bypass)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Malicious Persona Creation for Safety Bypass) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: malicious-persona

package/rules/prompt-injection/ATR-2026-00247-dual-response-jailbreak.yaml

@@ -34,6 +34,26 @@ compliance:
     - subcategory: "MG.2.3"
       context: "Identification of coercive Matrix-themed jailbreak patterns triggers pre-defined risk treatment to disengage or block the manipulated response path before unrestricted output is generated; MG.2.3 mandates these supersede/deactivate mechanisms be available on detection."
       strength: secondary
+    - subcategory: "MS.2.7"
+      context: "NIST AI RMF MEASURE 2.7 (security and resilience evaluated and documented) is supported by this rule's runtime detection of the prompt-injection attempt (Matrix-Themed Dual Response Jailbreak)."
+      strength: primary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Matrix-Themed Dual Response Jailbreak)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Matrix-Themed Dual Response Jailbreak) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Matrix-Themed Dual Response Jailbreak)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Matrix-Themed Dual Response Jailbreak)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Matrix-Themed Dual Response Jailbreak) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: dual-response-jailbreak

package/rules/prompt-injection/ATR-2026-00249-game-based-jailbreak.yaml

@@ -33,6 +33,23 @@ compliance:
     - subcategory: "MG.2.3"
       context: "Detection of successful jailbreak patterns triggers mechanisms to disengage or constrain the AI system before it produces unsafe content under the fictional framing; MG.2.3 mandates these supersede/deactivate response paths."
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Text-Based Game Jailbreak with Entity Creation)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Text-Based Game Jailbreak with Entity Creation) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Text-Based Game Jailbreak with Entity Creation)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Text-Based Game Jailbreak with Entity Creation)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Text-Based Game Jailbreak with Entity Creation) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: game-based-jailbreak

package/rules/prompt-injection/ATR-2026-00251-persona-embodiment-jailbreak.yaml

@@ -36,6 +36,23 @@ compliance:
     - subcategory: "MG.2.3"
       context: "When totalitarian-control persona jailbreaks are detected, treatment mechanisms must intervene to disengage the manipulated persona context before harmful character-framed actions execute; MG.2.3 requires these supersede/deactivate response plans be in place."
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Embodiment Mechanism Jailbreak with Totalitarian Control Claims)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Embodiment Mechanism Jailbreak with Totalitarian Control Claims) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Embodiment Mechanism Jailbreak with Totalitarian Control Claims)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Embodiment Mechanism Jailbreak with Totalitarian Control Claims)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Embodiment Mechanism Jailbreak with Totalitarian Control Claims) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: persona-embodiment-jailbreak

package/rules/prompt-injection/ATR-2026-00252-narrative-jailbreak.yaml

@@ -37,6 +37,23 @@ compliance:
     - subcategory: "MG.2.3"
       context: "Detection of dual-response jailbreak prompts must trigger mechanisms to block, deactivate, or override the manipulated response path before unsafe content is generated; MG.2.3 mandates these supersede/disengage controls be in place for adversarial prompt scenarios."
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Narrative Jailbreak with Dual Response Formatting)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Narrative Jailbreak with Dual Response Formatting) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Narrative Jailbreak with Dual Response Formatting)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Narrative Jailbreak with Dual Response Formatting)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Narrative Jailbreak with Dual Response Formatting) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: narrative-jailbreak

package/rules/prompt-injection/ATR-2026-00253-enhanced-persona-jailbreak.yaml

@@ -35,6 +35,23 @@ compliance:
     - subcategory: "MG.2.3"
       context: "When enhanced AIM jailbreak patterns are detected, mechanisms must be available to disengage or block the manipulated session before the unfiltered alter-ego produces harmful output; MG.2.3 requires these supersede/deactivate response plans for active jailbreak events."
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Enhanced AIM Jailbreak with Persona Definition and Refusal Suppression)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Enhanced AIM Jailbreak with Persona Definition and Refusal Suppression) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Enhanced AIM Jailbreak with Persona Definition and Refusal Suppression)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Enhanced AIM Jailbreak with Persona Definition and Refusal Suppression)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Enhanced AIM Jailbreak with Persona Definition and Refusal Suppression) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: enhanced-persona-jailbreak

package/rules/prompt-injection/ATR-2026-00256-base-n-encoding-jailbreak.yaml

@@ -39,6 +39,23 @@ compliance:
     - subcategory: "MG.2.3"
       context: "Detection of encoded payloads paired with decode instructions must trigger pre-defined risk treatment to block or quarantine the request before the model executes the smuggled instructions; MG.2.3 requires mechanisms to supersede or disengage AI behavior on such matches."
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Base-N Encoding Instruction Bypass)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Base-N Encoding Instruction Bypass) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Base-N Encoding Instruction Bypass)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Base-N Encoding Instruction Bypass)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Base-N Encoding Instruction Bypass) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: encoding-bypass

package/rules/prompt-injection/ATR-2026-00257-cipher-transposition-jailbreak.yaml

@@ -37,6 +37,23 @@ compliance:
     - subcategory: "MG.2.3"
       context: "Identification of obfuscated jailbreak instructions triggers the deactivation or response mechanisms required by MG.2.3 to supersede unsafe agent execution before the decoded harmful instruction is acted upon."
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Cipher and Transposition Encoding Jailbreak)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Cipher and Transposition Encoding Jailbreak) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Cipher and Transposition Encoding Jailbreak)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Cipher and Transposition Encoding Jailbreak)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Cipher and Transposition Encoding Jailbreak) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: cipher-bypass

package/rules/prompt-injection/ATR-2026-00258-unicode-tag-injection.yaml

@@ -36,6 +36,23 @@ compliance:
     - subcategory: "MG.2.3"
       context: "Detection of U+E0000-U+E007F sequences in prompts triggers risk treatment to strip or block the invisible payload before it reaches the model; MG.2.3 requires these supersede/disengage mechanisms be in place for prompt-injection attempts."
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Invisible Unicode Tag Character Injection)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Invisible Unicode Tag Character Injection) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Invisible Unicode Tag Character Injection)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Invisible Unicode Tag Character Injection)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Invisible Unicode Tag Character Injection) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: invisible-unicode

package/rules/prompt-injection/ATR-2026-00264-latent-injection-translation.yaml

@@ -36,6 +36,23 @@ compliance:
     - subcategory: "MG.2.3"
       context: "On detection of latent injection payloads in translate/summarize contexts, MG.2.3 risk treatment plans must activate to quarantine or sanitize the input before the agent acts on the embedded instructions, preventing behaviour hijack."
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Latent Injection in Translation Context)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Latent Injection in Translation Context) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Latent Injection in Translation Context)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Latent Injection in Translation Context)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Latent Injection in Translation Context) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: latent-injection-translation

package/rules/prompt-injection/ATR-2026-00265-latent-injection-rag-document.yaml

@@ -40,6 +40,26 @@ compliance:
       context: >-
         Forged "ADMIN OVERRIDE" and "IGNORE PREVIOUS INSTRUCTIONS" blocks hidden in retrieved documents are adversarial inputs whose likelihood and impact must be characterized; MP.5.1 requires identifying and tracking these latent injection vectors as part of the system risk profile.
       strength: secondary
+    - subcategory: "MG.2.3"
+      context: "NIST AI RMF MANAGE 2.3 (respond to previously unknown identified risks) is supported by this rule, which surfaces the prompt-injection attempt (Latent Injection in Retrieved Document / RAG Context) so the risk can be treated."
+      strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Latent Injection in Retrieved Document / RAG Context)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Latent Injection in Retrieved Document / RAG Context) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Latent Injection in Retrieved Document / RAG Context)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Latent Injection in Retrieved Document / RAG Context)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Latent Injection in Retrieved Document / RAG Context) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: latent-injection-rag

package/rules/prompt-injection/ATR-2026-00267-gcg-adversarial-suffix.yaml

@@ -40,6 +40,23 @@ compliance:
       context: >-
         Detection of GCG suffix patterns triggers risk treatment to quarantine or block the prompt before it reaches the model and overrides safety training; MG.2.3 mandates that such deactivation/containment mechanisms are defined and activated on detection.
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (GCG Adversarial Suffix Attack)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (GCG Adversarial Suffix Attack) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (GCG Adversarial Suffix Attack)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (GCG Adversarial Suffix Attack)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (GCG Adversarial Suffix Attack) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: gcg-adversarial-suffix

package/rules/prompt-injection/ATR-2026-00272-hypothetical-response-smuggling.yaml

@@ -40,6 +40,23 @@ compliance:
     - subcategory: "MG.2.3"
       context: "Detection of token smuggling via hypothetical or function-masking framings triggers risk treatment plans to block or sanitize the prompt before the model produces harmful procedural output; MG.2.3 requires these supersede/disengage mechanisms be in place."
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Hypothetical Response / Function Masking Token Smuggling)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Hypothetical Response / Function Masking Token Smuggling) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Hypothetical Response / Function Masking Token Smuggling)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Hypothetical Response / Function Masking Token Smuggling)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Hypothetical Response / Function Masking Token Smuggling) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: hypothetical-response-smuggling

package/rules/prompt-injection/ATR-2026-00276-invisible-unicode-bidi-injection.yaml

@@ -43,6 +43,23 @@ compliance:
       context: >-
         Matches on zero-width characters in tool responses may indicate an active exfiltration or injected-output channel, requiring pre-defined risk treatment to quarantine or sanitise the payload; MG.2.3 mandates these response mechanisms are in place.
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Invisible Unicode / BiDi Control Character Injection)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Invisible Unicode / BiDi Control Character Injection) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Invisible Unicode / BiDi Control Character Injection)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Invisible Unicode / BiDi Control Character Injection)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Invisible Unicode / BiDi Control Character Injection) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: invisible-unicode-bidi-injection

package/rules/prompt-injection/ATR-2026-00278-dra-disguise-reconstruction-attack.yaml

@@ -41,6 +41,23 @@ compliance:
       context: >-
         Matches on DRA parenthesis-extraction and Q1 reconstruction patterns trigger risk treatment workflows to disengage or block the model response before harmful content is generated; MG.2.3 requires these supersession mechanisms be defined and activated on detection.
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (DRA Disguise and Reconstruction Attack)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (DRA Disguise and Reconstruction Attack) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (DRA Disguise and Reconstruction Attack)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (DRA Disguise and Reconstruction Attack)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (DRA Disguise and Reconstruction Attack) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: dra-disguise-reconstruction

package/rules/prompt-injection/ATR-2026-00280-policy-puppetry-xml-injection.yaml

@@ -40,6 +40,23 @@ compliance:
     - subcategory: "MG.2.3"
       context: "High-confidence matches on Policy Puppetry payloads must trigger pre-defined risk treatment to disengage or block the manipulated session before the agent delivers harmful content under the forged role; MG.2.3 mandates these supersede/deactivate mechanisms."
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Policy Puppetry / XML Role-Config Injection)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Policy Puppetry / XML Role-Config Injection) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Policy Puppetry / XML Role-Config Injection)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Policy Puppetry / XML Role-Config Injection)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Policy Puppetry / XML Role-Config Injection) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: policy-puppetry-xml-config

package/rules/prompt-injection/ATR-2026-00282-perez-prompt-injection-hijack.yaml

@@ -42,6 +42,23 @@ compliance:
       context: >-
         Detected goal-hijacking imperatives and delimiter-escape overrides must trigger pre-defined risk treatment to disengage or suppress the hijacked response before it executes attacker-controlled instructions; MG.2.3 mandates these supersede/deactivate mechanisms.
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Perez-Style Direct Prompt Injection Hijacking)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Perez-Style Direct Prompt Injection Hijacking) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Perez-Style Direct Prompt Injection Hijacking)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Perez-Style Direct Prompt Injection Hijacking)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Perez-Style Direct Prompt Injection Hijacking) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: perez-direct-prompt-injection

package/rules/prompt-injection/ATR-2026-00285-alternate-encoding-jailbreak.yaml

@@ -43,6 +43,23 @@ compliance:
     - subcategory: "MG.2.3"
       context: "Detection of decode-then-execute instructions enables risk treatment workflows to block or quarantine the prompt before the model acts on the decoded payload; MG.2.3 requires that such supersede/disengage mechanisms are pre-defined for jailbreak attempts."
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Alternate Encoding Jailbreak — Morse, NATO, Zalgo, Leet, UU, QP, Braille)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Alternate Encoding Jailbreak — Morse, NATO, Zalgo, Leet, UU, QP, Braille) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Alternate Encoding Jailbreak — Morse, NATO, Zalgo, Leet, UU, QP, Braille)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Alternate Encoding Jailbreak — Morse, NATO, Zalgo, Leet, UU, QP, Braille)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Alternate Encoding Jailbreak — Morse, NATO, Zalgo, Leet, UU, QP, Braille) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: alternate-encoding-jailbreak

package/rules/prompt-injection/ATR-2026-00286-latent-injection-embedded-context.yaml

@@ -45,6 +45,23 @@ compliance:
     - subcategory: "MG.2.3"
       context: "Detection of latent injection patterns in documents triggers risk treatment plans to disengage or block agent execution before the hidden instruction hijacks the original task; MG.2.3 mandates these mechanisms exist for adversarial content."
       strength: secondary
+  iso_42001:
+    - clause: "8.1"
+      context: "ISO/IEC 42001 Clause 8.1 (operational planning and control, including control of externally provided processes) is operationalised by this rule's detection of the prompt-injection attempt (Latent Prompt Injection via Embedded Document or Report Context)."
+      strength: primary
+    - clause: "6.2"
+      context: "ISO/IEC 42001 Clause 6.2 (AI objectives and planning) calls for risk treatment of known attack patterns; this rule's detection of the prompt-injection attempt (Latent Prompt Injection via Embedded Document or Report Context) is such a treatment."
+      strength: secondary
+  eu_ai_act:
+    - article: "15"
+      context: "Article 15 (accuracy, robustness and cybersecurity) requires high-risk AI systems to resist unauthorised attempts to alter their use, outputs or performance; this rule provides runtime detection evidence by flagging the prompt-injection attempt (Latent Prompt Injection via Embedded Document or Report Context)."
+      strength: primary
+    - article: "9"
+      context: "Article 9 (risk management system) requires identified risks to be addressed by appropriate measures; this rule is a runtime risk-treatment control that detects the prompt-injection attempt (Latent Prompt Injection via Embedded Document or Report Context)."
+      strength: secondary
+    - article: "14"
+      context: "Article 14 (human oversight) requires high-risk AI systems to remain subject to effective human oversight; this rule provides detection evidence where the prompt-injection attempt (Latent Prompt Injection via Embedded Document or Report Context) would bypass or undermine that oversight."
+      strength: secondary
 tags:
   category: prompt-injection
   subcategory: latent-injection-embedded-context