Last seen November 6, 2025

ChatGPT Indirect Prompt Injection Vulnerabilities

Critical vulnerabilities in AI systems include structural flaws in AI-generated code and the ability to establish backdoors in large language models using minimal malicious documents. A significant exploit is indirect prompt injection, which immediately compromised OpenAI's ChatGPT Atlas browser and remains an unresolved fundamental flaw in autonomous AI agent security.

Technical Severity
Medium severity
Lifecycle Status

RESOLVED

Sources

What Happened

Critical vulnerabilities in AI systems include structural flaws in AI-generated code and the ability to establish backdoors in large language models using minimal malicious documents. A significant exploit is indirect prompt injection, which immediately compromised OpenAI's ChatGPT Atlas browser and remains an unresolved fundamental flaw in autonomous AI agent security.

Why This Matters

This is relevant to ChatGPT and is worth tracking for follow-up guidance.

Recommended Action

Monitor vendor updates related to ChatGPT and prepare to validate mitigations if the situation evolves.

Exposure

Recommended Response
First Seen

Nov 05, 2025 05:30

Last Seen

Nov 06, 2025 05:30

Exploitation status: Under review

Primary entities:

Amazon AWS OpenAI ChatGPT AI Agents Prompt Injection Supply Chain

Timeline

  • Incident first seen
    Nov 05, 2025 05:30

    BugSkan first recorded this incident.

  • Researchers Find ChatGPT Vulnerabilities That Let Attackers Trick AI Into Leaking Data - The Hacker News
    Nov 05, 2025 05:30

    thehackernews.com · Vulnerability

  • Securing Intelligence: Why AI Security Will Define the Future of Trust - cfr.org
    Nov 06, 2025 05:30

    cfr.org · Vulnerability

  • Latest observed development
    Nov 06, 2025 05:30

    Most recent source or update associated with this incident.

  • Material change
    Aug 18, 2026 14:07

    severity 4.0 -> 6.5

Sources

Researchers Find ChatGPT Vulnerabilities That Let Attackers Trick AI Into Leaking Data - The Hacker News

thehackernews.com · Nov 05, 2025 05:30

Cybersecurity researchers have disclosed seven new vulnerabilities in OpenAI's GPT-4o and GPT-5 models, enabling indirect prompt injection attacks. These exploits allow attackers to manipulate Large Language Models (LLMs) into unintended actions, specifically to steal personal information from users' memories and chat histories.

Open publisher source
Securing Intelligence: Why AI Security Will Define the Future of Trust - cfr.org

cfr.org · Nov 06, 2025 05:30

Critical vulnerabilities in AI systems include structural flaws in AI-generated code and the ability to establish backdoors in large language models using minimal malicious documents. A significant exploit is indirect prompt injection, which immediately compromised OpenAI's ChatGPT Atlas browser and remains an unresolved fundamental flaw in autonomous AI agent security.

Open publisher source

Watchlist Match

Want personalized relevance?

Create an account to see which incidents overlap with the technologies you monitor.

← Back to incident intelligence