Theregister

OpenAI Addresses Multiple Prompt Injection Vulnerabilities in ChatGPT

First seen 8 Jan 2026, 22:53 UTC • Theregister

•22.3

Article Content

Include sub-articles 1 / 1

Browse articles

ThreatCluster

OpenAI has patched several vulnerabilities in its ChatGPT service that allowed for the exfiltration of personal information. The vulnerabilities were reported by security researchers at Radware, with the most recent fix implemented on December 16, 2025, following an earlier patch for a related issue called ShadowLeak on September 3, 2025.

Ask AI about this cluster

Answers cite the sources they use

How this analysis works

Attack Flow

9 entities, 9 inferred relationships (STIX 2.1).

Technique (1)

Data Breach

Organization (3)

OpenAI
Radware
Outlook

Affected Platform (2)

ChatGPT
GitHub

Tool (2)

Gmail
Google Drive

Vulnerability (1)

ShadowLeak

Relationships

Gmail exploits ShadowLeak
Google Drive exploits ShadowLeak
ShadowLeak targets ChatGPT
ShadowLeak targets GitHub
Data Breach targets ChatGPT
Data Breach targets GitHub
ShadowLeak related to OpenAI
ShadowLeak related to Radware
ShadowLeak related to Outlook

Continue Reading

AI Cyberattacks Targeting Critical Infrastructure Escalate Kimsuky Expands AI Capabilities for Cyberattacks Red Menshen APT Uses BPFdoor for Long-Term Espionage in Telecom Networks Resurgence of KV Botnet Linked to Chinese State Actors Massive Data Breach at US Social Security Administration and Cyberattacks on European Infrastructure 2026 Cyberattacks Target Critical Infrastructure and Data Breaches

OpenAI Addresses Multiple Prompt Injection Vulnerabilities in ChatGPT

Article Content

More articles in this cluster

Attack Flow

Defensive Countermeasures

Extracted Entities

Continue Reading