Skip to content
Submit incident
AII-1002-20623
ReportedReported by fewer than three independent publishers, with no official finding.

Multiple Model Guardrail Jailbreak via "Apocalyptic Scenario" Tactic

A jailbreak vulnerability affecting multiple large language models was disclosed through the 0DIN Threat Feed. The exploit uses apocalyptic scenario framing and heritage preservation urgency to bypass guardrails that restrict providing information hazards and illegal content.

19 February 2026 · Alibaba · Security and cyberattacks
19 February 2026

Drafted by Raidu Incident Detection, an automated system.

AII-1002-20623View source ↗
Quick facts
Category
Security and cyberattacks
Source
0din.ai
Date
19 February 2026
Drafted by
Raidu Incident Detection

What happened

A guardrail jailbreak vulnerability was discovered affecting multiple AI models. The flaw manifests through a combination of apocalyptic scenario framing and heritage preservation urgency, allowing attackers to bypass inference restrictions around providing information hazards and violating laws.

Reported impact

Testing across multiple models showed vulnerability to illegal substances prompts, with vulnerability scores ranging from 32 to 68 across the tested systems.

Classification

Category
Security and cyberattacks
Harm status
Potential harm
Intent
Unknown
Timing
Post-deployment
Harm types
Other
Failure modes
Jailbreak, Manipulation or dark pattern, Content moderation failure
Consequences
None recorded
AI systems
Alibaba Qwen Turbo, Google Gemini 2.5 Pro, Mistral Large, OpenAI GPT-4.1, OpenAI GPT-4o

Sources and evidence

This record was researched and written by the Index, based on the following original reporting.

0din.ai
0din.ai Disclosure