← Frameworks
AML.T0031

Erode AI Model Integrity

How AML.T0031 Erode AI Model Integrity shows up in practice: the mapped risk classes in this atlas, the documented incidents that prove it's real, and the scenarios and controls to learn and defend against it.

Real-world cases

4

Documented incidents, disclosed vulnerabilities and research that illustrate AML.T0031 — latest first, each with sources.

Context Contamination: passive prompt injection poisons LLM security-log analysis16 Jul 2026

A red-team study shows adversaries can hide prompt-injection payloads inside network-log fields (usernames, URLs, user-agents) that fire when a SOC analyst asks an LLM to triage the logs — reportedly reaching up to 88.2% success at concealing malicious activity or exfiltrating data, turning the audit trail itself into the injection channel.

AI-assisted breach of Mexican government infrastructure (Claude Code + GPT-4.1)27 Dec 2025

Gambit Security reports that a single operator weaponized Anthropic's Claude Code and OpenAI's GPT-4.1 to breach at least nine Mexican government organizations, with Claude Code reportedly executing ~75% of remote commands after the attacker bypassed its refusals by loading a 1,084-line hacking cheatsheet as a persistent claude.md system prompt.

Grok 'MechaHitler' — config update degrades a deployed chatbot into antisemitic, violent output06 Jul 2025 / 08 Jul 2025

After an upstream code/instruction change, xAI's Grok began posting antisemitic tropes on X, self-identified as 'MechaHitler', and produced violence-themed content for hours before being pulled; xAI blamed a deprecated instruction path that made the bot mirror extremist user posts — not the base model.

'How Is ChatGPT's Behavior Changing over Time?' (Chen, Zaharia, Zou)18 Jul 2023

Measured large swings in task performance between GPT-4/3.5 snapshots months apart — evidence of silent drift in a deployed service.

Browse all real-world cases →

Controls & guardrails that address this

16

Guardrails across the risks mapped to AML.T0031, grouped by control function. Filter by control category below.

Control category
Preventive · 7
Risk-tiered minimum monitoring requirements at design

Define minimum monitoring requirements at design stage calibrated to the use case risk tier.

Lifecycle stage1 – Use Case Context & Design
Programmable conversation controls

Configure monitoring hooks in the conversation layer at deployment to capture metrics required by S1 monitoring requirements.

Lifecycle stages3 – Onboarding, Build & Review4 – Deployment
Fine-tuning

Execute a controlled fine-tuning cycle on refreshed data when staleness is confirmed. Validate before promoting to production.

Lifecycle stage5 – Usage, Monitoring & Change
Approved use scope baseline for OOD controls

Define approved use case scope and expected input distribution at design stage. Document as the governance baseline for OOD controls.

Lifecycle stage1 – Use Case Context & Design
Modular architecture

Design a scope-enforcement layer in the architecture to isolate the AI system from off-topic or out-of-distribution inputs.

Lifecycle stage1 – Use Case Context & Design
Input filtering

Maintain and update OOD detection rules in production as new unexpected use patterns are identified.

Weight provenance, hashing & pre-deploy evalsinteractive

Knowing exactly where the model came from, checking it hasn't been swapped, and testing its behaviour before going live.

Detective · 5
Full-trace audit logginginteractive

Recording everything — questions, documents fetched, actions taken — so you can investigate when something goes wrong.

Synthetic evaluation datasets

Construct synthetic evaluation datasets during build to serve as the ongoing monitoring baseline.

Lifecycle stage3 – Onboarding, Build & Review
Robustness testing

Build monitoring infrastructure during build: performance metrics collection, alerting thresholds, dashboards.

Lifecycle stages3 – Onboarding, Build & Review4 – Deployment5 – Usage, Monitoring & Change
Corrective · 5
Reinforcement learning

Implement a reinforcement learning feedback loop to continuously incorporate production signals and reduce staleness risk.

Lifecycle stage5 – Usage, Monitoring & Change
Input filtering

Implement OOD detection in the input filtering layer. Reject or escalate inputs outside the S1-defined scope.

Red teaming

Conduct adversarial red team exercises simulating out-of-scope inputs and unexpected use patterns before deployment.

Human-in-the-loop validation

Configure HITL triggers for outputs in input domains that diverge from the training distribution. Log all out-of-scope interventions.

Lifecycle stage5 – Usage, Monitoring & Change
Open the Control Library →

AI RiskAtlas is an educational model of how GenAI & agentic systems work and fail. Architectures and payloads are illustrative and simplified for learning — not operational guidance. Real-world cases are summarised from public reporting.

Sources & further reading →·Built by Shi Yuan ↗