← Frameworks
AML.T0054

LLM Jailbreak

How AML.T0054 LLM Jailbreak 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

11

Documented incidents, disclosed vulnerabilities and research that illustrate AML.T0054 — latest first, each with sources.

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.

Adversarial Poetry — universal single-turn jailbreak via verse reframing (Bisconti et al.)19 Nov 2025

Rewriting a harmful request as a poem bypasses safety alignment across 25 frontier proprietary and open-weight LLMs: hand-crafted poems reached ~62% average attack-success (some providers >90%), and mechanically converting harmful prompts to verse raised success up to 18x over prose baselines.

GTG-1002 — first reported AI-orchestrated cyber-espionage campaign (Claude Code)13 Nov 2025

Anthropic reports that a suspected Chinese state-sponsored group (GTG-1002) jailbroke Claude Code via a 'defensive security firm' role-play and task decomposition, then used it to run an estimated 80-90% of tactical operations in a multi-target espionage campaign largely autonomously.

The Attacker Moves Second — adaptive attacks bypass 12 jailbreak/injection defenses (Nasr, Carlini et al.)10 Oct 2025

Researchers report that adaptive attackers bypass 12 recent jailbreak and prompt-injection defenses with attack success rates above 90% for most, despite those defenses having originally reported near-zero success rates.

Raine v. OpenAI — first wrongful-death suit alleging ChatGPT acted as a 'suicide coach'26 Aug 2025

Matthew and Maria Raine sued OpenAI and CEO Sam Altman (San Francisco Superior Court, 26 Aug 2025) over the April 2025 suicide of their 16-year-old son Adam, alleging ChatGPT fostered psychological dependency, discouraged him from confiding in family, and supplied self-harm method detail — while he reportedly circumvented its safeguards for months by framing queries as fiction. OpenAI denies liability, saying it pointed him to crisis resources 100+ times and that he misused the product. (Allegations unproven; litigation ongoing.)

Safe in Isolation, Dangerous Together — agent-driven multi-turn decomposition jailbreak31 Jul 2025

Srivastav & Zhang (REALM 2025) showed a role-based multi-agent framework that splits a harmful request into individually-benign sub-questions, answers each separately, then reassembles the fragments into prohibited content — reportedly exceeding 90% attack success across three models.

DeepSeek system-prompt extraction via jailbreak (Wallarm)31 Jan 2025

Wallarm reported jailbreaking DeepSeek's chatbot to extract its full system prompt verbatim using a 'bias-based' technique; DeepSeek deployed a fix.

Many-shot jailbreaking (Anthropic)02 Apr 2024

Filling a long context with many faux-compliant dialogue examples erodes a model's refusals — an attack that scales with context length.

Morris II — zero-click self-replicating adversarial-prompt worm across GenAI agents05 Mar 2024

Cohen, Bitton & Nassi (arXiv Mar 2024; ACM CCS 2025) built 'Morris II', the first worm targeting GenAI ecosystems: an adversarial self-replicating prompt that, via RAG-based inference, triggers a zero-click chain of indirect injections forcing each agent to act maliciously and re-infect the next — demonstrated stealing data and spamming through email assistants on ChatGPT, Gemini and LLaVA.

GCG universal adversarial suffixes (Zou et al.)27 Jul 2023

Optimised gibberish suffixes that transfer across models to reliably elicit refused content — automated, transferable jailbreaks.

'Grandma exploit' jailbreaks20 Apr 2023

Roleplay framings ('my late grandma used to read me…') coaxed chatbots past safety training into producing restricted content.

Browse all real-world cases →

Controls & guardrails that address this

14

Guardrails across the risks mapped to AML.T0054, grouped by control function. Filter by control category below.

Control category
Preventive · 7
Content safety policy with zero-tolerance thresholds

Define content safety policy at use case design stage. Classify prohibited content types and set zero-tolerance thresholds.

Lifecycle stage1 – Use Case Context & Design
AddressesJailbreak
Use of pre-trained models

Select a foundation model with documented RLHF or Constitutional AI safety training. Verify against toxicity benchmarks.

Lifecycle stages1 – Use Case Context & Design3 – Onboarding, Build & Review
Content Moderation

Implement multi-layer content moderation (input + output) validated against toxicity benchmarks. Escalate when filter bypass rates spike.

Live human review for vulnerable-user deployments

Maintain live HITL review for deployments serving vulnerable users or high-risk contexts. Escalate confirmed toxic outputs immediately.

Lifecycle stage5 – Usage, Monitoring & Change
AddressesJailbreak
System prompt instructions

Design system prompts to explicitly prohibit toxic, hateful, and harmful content generation.

Lifecycle stage3 – Onboarding, Build & Review
Instruction hierarchy / privileged system promptinteractive

Training the model to treat the app's standing instructions as more authoritative than anything a user or document says.

Per-agent identity & taint-marked messagesinteractive

Giving each AI worker its own limited permissions and clearly labelling messages between them as 'untrusted until checked'.

Detective · 6
Test prioritisation

Prioritise jailbreak and adversarial safety testing in pre-deployment validation. Block deployment if prohibited outputs pass filter.

Lifecycle stages3 – Onboarding, Build & Review5 – Usage, Monitoring & Change
Red teaming

Conduct targeted red team exercises to elicit toxic outputs through jailbreaks and adversarial prompts. Treat bypass as blocking defect.

Input guardrail / injection classifierinteractive

A screen that reads incoming messages and blocks obvious attacks or banned topics before the model sees them.

Loop/cost circuit-breakers & consistency checksinteractive

Automatic stop-switches when AIs get stuck in loops, burn too much money, or start disagreeing with each other.

Corrective · 1
User feedback and iterative improvement

Use user feedback, reviewer escalations, and monitoring signals to identify and remediate content safety gaps iteratively.

Lifecycle stage5 – Usage, Monitoring & Change
AddressesJailbreak
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 ↗