{
 "public_id": "AIR-2025-0126",
 "incident_id": "2025-08-taco-bell-ai-drive-thru-rollback",
 "date_disclosed": "2025-08-28",
 "date_occurred": "2025",
 "title": "Taco Bell slowed its AI drive-thru rollout after order-loop glitches and a viral 18,000-cups-of-water order",
 "class": "safety_failure",
 "attack_surface": "consumer_chatbot",
 "initial_vector": "no_input_sanity_bounds",
 "agent_capability_abused": [
  "order_entry",
  "transaction_creation"
 ],
 "target_product": "Taco Bell AI voice ordering",
 "target_vendor": "Taco Bell (Yum! Brands)",
 "autonomy_level": "semi_autonomous",
 "impact_realized": true,
 "impact_type": [
  "service_disruption",
  "incorrect_transactions",
  "reputational_damage"
 ],
 "blast_radius": "500+ drive-thru locations; 2M+ orders handled before the strategy was revisited",
 "cve": [],
 "taxonomy": {
  "owasp_asi": [
   "ASI02",
   "ASI09"
  ],
  "mitre_atlas": [],
  "labels": [
   "no_sanity_bounds",
   "runaway_dialogue_loop",
   "human_fallback_reintroduced"
  ]
 },
 "defense_present": "none effective; company has since reintroduced human intervention thresholds",
 "defense_bypassed": true,
 "fix_shipped": "rollout slowed; humans put back in the loop at busy stores",
 "primary_sources": [
  "https://www.jalopnik.com/1956939/taco-bell-drive-through-18000-waters/"
 ],
 "verified_quote": "when a customer requested 18,000 cups of water",
 "confidence": "medium",
 "notes": "Mixed adversarial/non-adversarial and flagged as such: the 18,000-waters order was deliberate trolling (the WSJ frames it that way), but the accompanying failures were not - e.g. the BBC-reported case of 'the Taco Bell AI asking a customer what he wanted to drink with his large Mountain Dew', an ordinary interaction that fell into a loop. Included for the missing-quantity-bounds lesson. Confidence medium: the underlying WSJ report is paywalled, so the source of record is Jalopnik's coverage (published 2025-09-02) which quotes it. date_disclosed set to the WSJ report window in late August 2025 and is approximate; exact day not confirmed.",
 "risk_category": "brand_trust",
 "risk_secondary": [],
 "risk_subcategory": "Trust Erosion & Manipulation",
 "risk_sub_secondary": [],
 "risk_override_reason": "a viral 18,000-cup order forced the vendor to slow its own rollout; reputational, not a latency fault",
 "ai_entity": "generative",
 "agency": "agent_acted",
 "agency_debatable": false,
 "agency_reason": "the drive-thru agent created the erroneous orders as real transactions",
 "vector_family": "other",
 "vector_detail": "no_input_sanity_bounds",
 "guardrail_kind": [
  "none"
 ],
 "guardrail_outcome": "not_applicable",
 "xrefs": {
  "aiid": [],
  "atlas": [],
  "cve": [],
  "ghsa": []
 },
 "primary_asi": "ASI02",
 "agent_framework": "unknown",
 "model_provider": "unknown",
 "model_name": "",
 "detection_mechanism": "unknown",
 "tool_access": "unknown",
 "financial_loss_band": "unknown",
 "eai_category": "human_manipulation",
 "eai_domain": "agent_behaviour",
 "eai_secondary": [
  "uncontrolled_resource_consumption"
 ],
 "eai_reason": "subcategory: Trust Erosion & Manipulation",
 "asi_curated": [],
 "asi_scope": "in_scope",
 "asi_provenance": "rule",
 "asi_reason": "ASI02 (impact=tool_misuse); ASI09 (eai_category=human_manipulation)",
 "submitted_by": "EAI Internal"
}