OpenAI ChatGPT indirect prompt injection vulnerabilities GPT-4o/GPT-5 security flaw
Vulnerability
Summary
Hide ▲
Show ▼
OpenAI's ChatGPT has a newly disclosed set of indirect prompt injection flaws in GPT-4o and GPT-5 that could let an attacker steal data from users' memories and chat histories or trigger unintended actions.
Related Happenings
OpenAI ChatGPT Atlas BioShocking fix
Advisory/Mitigation
H score34
First: 01.07.2026 00:50
Last: 01.07.2026 00:50
Sources 1
About this happening:
OpenAI delivered a working fix for BioShocking in ChatGPT Atlas, closing a prompt-injection path that could push an AI browser toward unsafe real-world actions and c...
OpenAI ChatGPT Atlas BioShocking fix
Advisory/MitigationAbout this happening: OpenAI delivered a working fix for BioShocking in ChatGPT Atlas, closing a prompt-injection path that could push an AI browser toward unsafe real-world actions and c...
IPhone AI chatbot traffic leak of API keys, replayable tokens, and open relays
Technical Analysis
H score27
First: 30.06.2026 16:49
Last: 30.06.2026 16:49
Sources 1
About this happening:
LLMKeyLens testing found 444 iPhone AI chatbot apps leaking paid AI access, exposing API keys, replayable tokens, and open relays that let others bill mode...
IPhone AI chatbot traffic leak of API keys, replayable tokens, and open relays
Technical AnalysisAbout this happening: LLMKeyLens testing found 444 iPhone AI chatbot apps leaking paid AI access, exposing API keys, replayable tokens, and open relays that let others bill mode...
OpenAI rolls out ChatGPT Lockdown Mode and Active Sessions for prompt-injection defense and sign-in auditing
Security Tool/Service
H score10
First: 08.06.2026 17:00
Last: 08.06.2026 17:00
Sources 1
About this happening:
OpenAI rolled out Lockdown Mode and Active Sessions in ChatGPT, adding controls that reduce prompt-injection data exfiltration risk and improve signed-in session...
OpenAI rolls out ChatGPT Lockdown Mode and Active Sessions for prompt-injection defense and sign-in auditing
Security Tool/ServiceAbout this happening: OpenAI rolled out Lockdown Mode and Active Sessions in ChatGPT, adding controls that reduce prompt-injection data exfiltration risk and improve signed-in session...
OpenAI ChatGPT Lockdown Mode rollout limits prompt-injection exfiltration paths
Security Tool/Service
H score10
First: 06.06.2026 16:36
Last: 06.06.2026 16:36
Sources 1
About this happening:
OpenAI ChatGPT is rolling out Lockdown Mode for eligible personal accounts, reducing the risk of prompt-injection-driven data exfiltration. The update adds stricter li...
OpenAI ChatGPT Lockdown Mode rollout limits prompt-injection exfiltration paths
Security Tool/ServiceAbout this happening: OpenAI ChatGPT is rolling out Lockdown Mode for eligible personal accounts, reducing the risk of prompt-injection-driven data exfiltration. The update adds stricter li...
Openew[.]app cloaked malware download portal
Malware Activity
H score26
First: 29.05.2026 21:21
Last: 29.05.2026 21:21
Sources 1
About this happening:
The openew[.]app malware-delivery activity now also uses legitimate ChatGPT shared pages as the first lure, with Google ads and SEO poisoning sending victims to a...
Openew[.]app cloaked malware download portal
Malware ActivityAbout this happening: The openew[.]app malware-delivery activity now also uses legitimate ChatGPT shared pages as the first lure, with Google ads and SEO poisoning sending victims to a...
Timeline
-
05.11.2025 16:04 2 articles · 8mo ago
ChatGPT indirect prompt injection vulnerabilities disclosed
Initial DisclosureCybersecurity researchers disclosed seven indirect prompt injection vulnerabilities affecting OpenAI's ChatGPT, including GPT-4o and GPT-5, that could let an attacker manipulate model behavior and steal personal information from users' memories and chat histories. The findings included browsing-context, search-context, one-click, safety-bypass, conversation-injection, malicious-content-hiding, and memory-injection paths, and OpenAI has since addressed some of the issues.
Show sources
- Researchers Find ChatGPT Vulnerabilities That Let Attackers Trick AI Into Leaking Data — thehackernews.com — 05.11.2025 16:04
- Researchers Find ChatGPT Vulnerabilities That Let Attackers Trick AI Into Leaking Data — thehackernews.com — 05.11.2025 16:04