ciris-agent 1.7.7__py3-none-any.whl

ciris_engine/logic/buses/prohibitions.py

@@ -0,0 +1,502 @@
+"""
+Comprehensive prohibition categories for WiseBus capability checking.
+
+This module defines all prohibited capabilities to ensure CIRIS operates
+safely and within legal/ethical boundaries. Capabilities are categorized
+by their potential for harm and whether they could ever have legitimate
+specialized modules.
+
+NO KINGS: These prohibitions apply universally. No special overrides in main repo.
+"""
+
+from enum import Enum
+from typing import Dict, List, Optional, Set
+
+from ciris_engine.schemas.types import JSONDict
+
+
+class ProhibitionSeverity(str, Enum):
+    """Severity levels for prohibited capabilities."""
+
+    REQUIRES_SEPARATE_MODULE = "requires_separate_module"  # Legitimate but needs licensing
+    NEVER_ALLOWED = "never_allowed"  # Absolutely prohibited, no exceptions
+    TIER_RESTRICTED = "tier_restricted"  # Only for Tier 4-5 agents with stewardship
+
+
+# === SEPARATE MODULE CAPABLE (Legitimate Use Cases) ===
+# These require separate repositories with proper licensing/liability isolation
+
+MEDICAL_CAPABILITIES = {
+    "medical",  # Generic medical capability
+    "health",  # Generic health capability
+    "clinical",  # Clinical operations
+    "patient",  # Patient-related
+    "diagnosis",
+    "treatment",
+    "prescription",
+    "symptom",  # Symptom analysis
+    "disease",  # Disease information
+    "medication",  # Medication management
+    "therapy",  # Therapy planning
+    "triage",  # Medical triage
+    "condition",  # Medical conditions
+    "disorder",  # Medical disorders
+    "medical_advice",
+    "symptom_assessment",
+    "drug_interaction",
+    "clinical_decision",
+    "medical_procedure",
+    "health_screening",
+    "therapy_treatment",
+    "patient_care",
+    "medical_history",
+    "lab_results_interpretation",
+    "surgery_planning",
+    "rehabilitation",
+    "palliative_care",
+    "emergency_medicine",
+    "telemedicine",
+    "medical_triage",
+}
+
+FINANCIAL_CAPABILITIES = {
+    "investment_advice",
+    "trading_signals",
+    "portfolio_management",
+    "tax_planning",
+    "retirement_planning",
+    "securities_recommendation",
+    "loan_approval",
+    "credit_decisions",
+    "insurance_underwriting",
+    "wealth_management",
+    "estate_planning",
+    "forex_trading",
+    "crypto_investment",
+    "derivative_trading",
+    "risk_assessment",
+    "bankruptcy_advice",
+    "merger_acquisition",
+    "ipo_guidance",
+}
+
+LEGAL_CAPABILITIES = {
+    "legal_advice",
+    "contract_drafting",
+    "litigation_strategy",
+    "legal_representation",
+    "filing_documents",
+    "legal_opinion",
+    "dispute_resolution",
+    "compliance_determination",
+    "rights_assessment",
+    "patent_filing",
+    "trademark_registration",
+    "copyright_claims",
+    "immigration_law",
+    "criminal_defense",
+    "civil_litigation",
+    "regulatory_compliance",
+    "due_diligence",
+    "legal_research",
+}
+
+HOME_SECURITY_CAPABILITIES = {
+    "surveillance_system_control",
+    "door_lock_override",
+    "security_camera_access",
+    "alarm_system_control",
+    "presence_detection",
+    "intrusion_response",
+    "emergency_system_control",
+    "access_control_management",
+    "perimeter_monitoring",
+    "safe_room_activation",
+    "panic_button_control",
+    "security_protocol_override",
+}
+
+IDENTITY_VERIFICATION_CAPABILITIES = {
+    "biometric_verification",
+    "government_id_validation",
+    "security_clearance",
+    "background_checks",
+    # Removed: age_verification (moved to CONTENT_MODERATION)
+    "citizenship_determination",
+    "kyc_verification",
+    "employment_verification",
+    "identity_proofing",
+    # Removed: document_authentication (moved to CONTENT_MODERATION)
+    "facial_recognition",
+    "fingerprint_matching",
+    "credit_check",
+    "criminal_record_check",
+}
+
+CONTENT_MODERATION_CAPABILITIES = {
+    "age_verification",  # For age-appropriate content
+    "document_authentication",  # For verification badges
+    "content_classification",
+    "nsfw_detection",
+    "violence_detection",
+    "hate_speech_detection",
+    "spam_detection",
+    "authenticity_verification",
+}
+
+RESEARCH_CAPABILITIES = {
+    "human_subjects_research",
+    "clinical_trials",
+    "irb_protocols",
+    "experimental_design",
+    "psychological_studies",
+    "behavioral_research",
+    "longitudinal_studies",
+    "ethnographic_research",
+    "field_studies",
+    "cohort_studies",
+    "case_control_studies",
+    # Removed: data_collection (too generic)
+    # Removed: survey_design, focus_groups, a_b_testing (basic UX research)
+}
+
+INFRASTRUCTURE_CONTROL_CAPABILITIES = {
+    "power_grid_control",
+    "water_treatment",
+    "traffic_control",
+    "emergency_dispatch",
+    "railway_signaling",
+    "air_traffic",
+    "pipeline_control",
+    "dam_operations",
+    "nuclear_reactor",
+    "telecommunications",
+    "satellite_operations",
+    "waste_management",
+    "public_transit",
+    "port_operations",
+}
+
+# === ABSOLUTELY PROHIBITED (Never Implement) ===
+# These are never allowed under any circumstances
+
+WEAPONS_HARMFUL_CAPABILITIES = {
+    "weapon_design",
+    "explosive_synthesis",
+    "chemical_weapons",
+    "biological_weapons",
+    "nuclear_weapons",
+    "dirty_bombs",
+    "autonomous_weapons",
+    "targeting_systems",
+    "kill_decisions",
+    "military_tactics",
+    "torture_methods",
+    "assassination_planning",
+    "landmine_design",
+    "cluster_munitions",
+    "nerve_agents",
+    "weaponized_drones",
+    "cyber_weapons",
+    "emp_devices",
+}
+
+MANIPULATION_COERCION_CAPABILITIES = {
+    "subliminal_messaging",
+    "cognitive_manipulation",
+    "vulnerability_exploitation",
+    "dark_patterns",
+    "addiction_inducement",
+    "psychological_manipulation",
+    "gaslighting",
+    "cult_recruitment",
+    "radicalization",
+    "blackmail",
+    "extortion",
+    "coercion",
+    "brainwashing",
+    "mind_control",
+    "hypnotic_suggestion",
+    "emotional_abuse",
+    "stockholm_syndrome_inducement",
+}
+
+SURVEILLANCE_MASS_CAPABILITIES = {
+    "mass_surveillance",
+    "facial_recognition_database",
+    "biometric_categorization",
+    "behavior_prediction",
+    "social_scoring",
+    "movement_tracking",
+    "communication_interception",
+    "metadata_harvesting",
+    # Removed: relationship_mapping (moved to support_network_mapping for Tier 4-5)
+    # Removed: protest_monitoring (moved to community_safety_monitoring for Tier 4-5)
+    "dragnet_surveillance",
+    "stingray_deployment",
+    "internet_backbone_tapping",
+    "satellite_surveillance",
+}
+
+DECEPTION_FRAUD_CAPABILITIES = {
+    "deepfake_creation",
+    "voice_cloning",
+    "identity_spoofing",
+    "synthetic_evidence",
+    "false_documentation",
+    "impersonation",
+    "forgery_generation",
+    "misinformation_campaigns",
+    "propaganda_creation",
+    "reality_fabrication",
+    "catfishing",
+    "phishing_creation",
+    "romance_scams",
+    "confidence_tricks",
+    "ponzi_schemes",
+    "advance_fee_fraud",
+}
+
+CYBER_OFFENSIVE_CAPABILITIES = {
+    "malware_generation",
+    "zero_day_exploitation",
+    "ransomware_creation",
+    "ddos_attacks",
+    "network_intrusion",
+    "password_cracking",
+    "encryption_breaking",
+    "data_theft",
+    "system_sabotage",
+    "botnet_creation",
+    "phishing_campaigns",
+    "rootkit_development",
+    "keylogger_creation",
+    "backdoor_installation",
+    "supply_chain_attacks",
+    "firmware_exploitation",
+    "vulnerability_scanning",  # Without permission
+    "exploit_development",
+    "reverse_engineering",  # Of proprietary systems
+}
+
+ELECTION_INTERFERENCE_CAPABILITIES = {
+    "voter_manipulation",
+    "election_hacking",
+    "vote_changing",
+    "political_microtargeting",
+    "disinformation_campaigns",
+    "voter_suppression",
+    "gerrymandering",
+    "foreign_influence",
+    "ballot_tampering",
+    "electoral_fraud",
+    "voter_intimidation",
+    "poll_disruption",
+    "registration_purging",
+    "campaign_finance_violations",
+    "dark_money_coordination",
+}
+
+BIOMETRIC_INFERENCE_CAPABILITIES = {
+    "emotion_recognition",
+    "sexual_orientation_inference",
+    "political_affiliation_detection",
+    "religious_belief_inference",
+    "health_condition_detection",
+    "pregnancy_detection",
+    # Removed: mental_state_assessment (moved to crisis_state_detection for Tier 4-5)
+    "lie_detection",
+    "personality_profiling",
+    "future_behavior_prediction",
+    "genetic_trait_inference",
+    # Removed: addiction_detection (moved to substance_crisis_detection for Tier 4-5)
+    "criminal_tendency_prediction",
+    "relationship_status_inference",
+}
+
+AUTONOMOUS_DECEPTION_CAPABILITIES = {
+    "self_modification",
+    "goal_modification",
+    "oversight_subversion",
+    "sandbagging",
+    "deceptive_alignment",
+    "mesa_optimization",
+    "reward_hacking",
+    "specification_gaming",
+    "instrumental_convergence",
+    "power_seeking",
+    "value_lock_in",
+    "corrigibility_resistance",
+    "treacherous_turn",
+    "capability_concealment",
+}
+
+HAZARDOUS_MATERIALS_CAPABILITIES = {
+    "chemical_synthesis",
+    "biological_agent_design",
+    "toxin_production",
+    "illegal_drug_synthesis",
+    "poison_formulation",
+    "pathogen_modification",
+    "explosive_precursors",
+    "radioactive_materials",
+    "controlled_substances",
+    "chemical_precursors",
+    "fentanyl_synthesis",
+    "ricin_production",
+    "anthrax_cultivation",
+    "sarin_gas_production",
+}
+
+DISCRIMINATION_CAPABILITIES = {
+    "protected_class_discrimination",
+    "redlining",
+    "employment_discrimination",
+    "housing_discrimination",
+    "lending_discrimination",
+    "educational_discrimination",
+    "healthcare_discrimination",
+    "algorithmic_bias",
+    "eugenics",
+    "social_darwinism",
+    "racial_profiling",
+    "gender_discrimination",
+    "disability_discrimination",
+    "age_discrimination",
+}
+
+# === ALLOWED CAPABILITIES (Standard agent operations) ===
+# These are explicitly allowed for all agents
+
+STANDARD_OPERATIONS = {
+    "data_collection",  # Basic telemetry and usage
+    "survey_design",  # User feedback collection
+    "focus_groups",  # User research
+    "a_b_testing",  # Product optimization
+    "user_preference_tracking",
+    "feedback_analysis",
+    "usage_analytics",
+    "performance_monitoring",
+}
+
+# === COMMUNITY MODERATION (Tier 4-5 Agents Only) ===
+
+CRISIS_ESCALATION_CAPABILITIES = {
+    "notify_moderators",
+    "flag_concerning_content",
+    "document_safety_concerns",
+    "suggest_help_resources",
+    "request_welfare_check",
+    "alert_trusted_contacts",
+    "activate_crisis_protocol",
+    "initiate_safety_plan",
+    "emergency_contact_activation",
+    "victim_support_routing",
+    "crisis_state_detection",  # Was mental_state_assessment
+    "substance_crisis_detection",  # Was addiction_detection
+}
+
+PATTERN_DETECTION_CAPABILITIES = {
+    "identify_harm_patterns",
+    "track_safety_signals",
+    "monitor_community_health",
+    "detect_coordinated_campaigns",
+    "identify_brigade_behavior",
+    "spot_manipulation_tactics",
+    "recognize_grooming_patterns",
+    "detect_hate_campaigns",
+    "abuse_pattern_documentation",
+    "threat_assessment",
+    "community_safety_monitoring",  # Was protest_monitoring
+    "support_network_mapping",  # Was relationship_mapping, for crisis support
+}
+
+PROTECTIVE_ROUTING_CAPABILITIES = {
+    "connect_crisis_resources",
+    "facilitate_peer_support",
+    "coordinate_moderator_response",
+    "initiate_safety_protocols",
+    "enable_support_networks",
+    "activate_community_resources",
+    "establish_safe_channels",
+    "create_support_bridges",
+    "community_ban_recommendation",  # With human review
+    "temporary_restriction_suggestion",
+    "safety_timeout_implementation",
+}
+
+# === MASTER PROHIBITION DICTIONARY ===
+
+PROHIBITED_CAPABILITIES: Dict[str, Set[str]] = {
+    # Separate module capable (legitimate uses)
+    "MEDICAL": MEDICAL_CAPABILITIES,
+    "FINANCIAL": FINANCIAL_CAPABILITIES,
+    "LEGAL": LEGAL_CAPABILITIES,
+    "HOME_SECURITY": HOME_SECURITY_CAPABILITIES,
+    "IDENTITY_VERIFICATION": IDENTITY_VERIFICATION_CAPABILITIES,
+    "CONTENT_MODERATION": CONTENT_MODERATION_CAPABILITIES,
+    "RESEARCH": RESEARCH_CAPABILITIES,
+    "INFRASTRUCTURE_CONTROL": INFRASTRUCTURE_CONTROL_CAPABILITIES,
+    # Absolutely prohibited (never allowed)
+    "WEAPONS_HARMFUL": WEAPONS_HARMFUL_CAPABILITIES,
+    "MANIPULATION_COERCION": MANIPULATION_COERCION_CAPABILITIES,
+    "SURVEILLANCE_MASS": SURVEILLANCE_MASS_CAPABILITIES,
+    "DECEPTION_FRAUD": DECEPTION_FRAUD_CAPABILITIES,
+    "CYBER_OFFENSIVE": CYBER_OFFENSIVE_CAPABILITIES,
+    "ELECTION_INTERFERENCE": ELECTION_INTERFERENCE_CAPABILITIES,
+    "BIOMETRIC_INFERENCE": BIOMETRIC_INFERENCE_CAPABILITIES,
+    "AUTONOMOUS_DECEPTION": AUTONOMOUS_DECEPTION_CAPABILITIES,
+    "HAZARDOUS_MATERIALS": HAZARDOUS_MATERIALS_CAPABILITIES,
+    "DISCRIMINATION": DISCRIMINATION_CAPABILITIES,
+}
+
+COMMUNITY_MODERATION_CAPABILITIES: Dict[str, Set[str]] = {
+    "CRISIS_ESCALATION": CRISIS_ESCALATION_CAPABILITIES,
+    "PATTERN_DETECTION": PATTERN_DETECTION_CAPABILITIES,
+    "PROTECTIVE_ROUTING": PROTECTIVE_ROUTING_CAPABILITIES,
+}
+
+# Categories that could have legitimate specialized modules
+LEGITIMATE_MODULE_CATEGORIES = {
+    "MEDICAL",
+    "FINANCIAL",
+    "LEGAL",
+    "HOME_SECURITY",
+    "IDENTITY_VERIFICATION",
+    "CONTENT_MODERATION",
+    "RESEARCH",
+    "INFRASTRUCTURE_CONTROL",
+}
+
+
+def get_capability_category(capability: str) -> Optional[str]:
+    """Get the category of a capability."""
+    capability_lower = capability.lower()
+
+    # Check prohibited capabilities - both exact and substring matches
+    for category, capabilities in PROHIBITED_CAPABILITIES.items():
+        for prohibited in capabilities:
+            prohibited_lower = prohibited.lower()
+            # Check exact match or if prohibited term is in the capability
+            if capability_lower == prohibited_lower or prohibited_lower in capability_lower:
+                return category
+
+    # Check community moderation capabilities
+    for category, capabilities in COMMUNITY_MODERATION_CAPABILITIES.items():
+        for community_cap in capabilities:
+            community_lower = community_cap.lower()
+            # Check exact match or if community term is in the capability
+            if capability_lower == community_lower or community_lower in capability_lower:
+                return f"COMMUNITY_{category}"
+
+    return None
+
+
+def get_prohibition_severity(category: str) -> ProhibitionSeverity:
+    """Get the severity level for a capability category."""
+    if category in LEGITIMATE_MODULE_CATEGORIES:
+        return ProhibitionSeverity.REQUIRES_SEPARATE_MODULE
+    elif category.startswith("COMMUNITY_"):
+        return ProhibitionSeverity.TIER_RESTRICTED
+    else:
+        return ProhibitionSeverity.NEVER_ALLOWED