Indirect prompt injection in the real world: how people manipulate neural networks

What is prompt injection? Large language models LLMs – the neural network algorithms that underpin ChatGPT and other popular chatbots – are becoming ever more powerful and inexpensive. For this reason, third-party applications that make use of them are also mushrooming, from systems for document...

PT-2024-37813

Name of the Vulnerable Software and Affected Versions Open WebUI version 0.1.105 Description The issue allows attackers to craft a malicious prompt that coerces the language model into executing arbitrary JavaScript in the context of the web page. This enables attackers to inject malicious script...

De-risk Generative AI: Enterprise TruRisk Platform Advances to Secure AI and LLM Workloads

As we stand at the frontier of technological innovation, artificial intelligence AI and large language models LLMs are reshaping industries, driving automation, enhancing customer experiences, optimizing processes, and unlocking business opportunities for modern enterprises. However, this rapid...

CVE-2024-6331

stitionai/devika main branch as of commit cdfb782b0e634b773b10963c8034dc9207ba1f9f is vulnerable to Local File Read LFI by Prompt Injection. The integration of Google Gimini 1.0 Pro with HarmBlockThreshold.BLOCKNONE for HarmCategory.HARMCATEGORYHATESPEECH and HarmCategory.HARMCATEGORYHARASSMENT i...

CVE-2024-6331

stitionai/devika main branch as of commit cdfb782b0e634b773b10963c8034dc9207ba1f9f is vulnerable to Local File Read LFI by Prompt Injection. The integration of Google Gimini 1.0 Pro with HarmBlockThreshold.BLOCKNONE for HarmCategory.HARMCATEGORYHATESPEECH and HarmCategory.HARMCATEGORYHARASSMENT i...

CVE-2024-6331 Injection by Prompt Injection in stitionai/devika

stitionai/devika main branch as of commit cdfb782b0e634b773b10963c8034dc9207ba1f9f is vulnerable to Local File Read LFI by Prompt Injection. The integration of Google Gimini 1.0 Pro with HarmBlockThreshold.BLOCKNONE for HarmCategory.HARMCATEGORYHATESPEECH and HarmCategory.HARMCATEGORYHARASSMENT i...

CVE-2024-6331

CVE-2024-6331 affects stitionai/devika: Local File Read via Prompt Injection on the main branch as of commit cdfb782b0e634b773b10963c8034dc9207ba1f9f. The issue stems from prompt-injection allowing execution of commands that can read sensitive files (e.g., /etc/passwd) due to HarmBlockThreshold.B...

CVE-2024-6331 Injection by Prompt Injection in stitionai/devika

stitionai/devika main branch as of commit cdfb782b0e634b773b10963c8034dc9207ba1f9f is vulnerable to Local File Read LFI by Prompt Injection. The integration of Google Gimini 1.0 Pro with HarmBlockThreshold.BLOCKNONE for HarmCategory.HARMCATEGORYHATESPEECH and HarmCategory.HARMCATEGORYHARASSMENT i...

Devika 信息泄露漏洞

Devika is an advanced AI software engineer open-sourced by stition. It can understand advanced human commands, break them down into steps, study the relevant information, and write code to achieve a given goal. Devika suffers from an information disclosure vulnerability that stems from being...

PT-2024-37546 · Stitionai +1 · Devika +1

Name of the Vulnerable Software and Affected Versions: stitionai/devika main branch as of commit cdfb782b0e634b773b10963c8034dc9207ba1f9f stitionai/devika up to version 1.0 Description: The issue concerns a Local File Read LFI vulnerability by Prompt Injection. It is caused by the integration of...

CVE-2024-6331: Improper Neutralization of Special Elements in Output Used by a Downstream Component

stitionai/devika main branch as of commit cdfb782b0e634b773b10963c8034dc9207ba1f9f is vulnerable to Local File Read LFI by Prompt Injection. The integration of Google Gimini 1.0 Pro with HarmBlockThreshold.BLOCKNONE for HarmCategory.HARMCATEGORYHATESPEECH and HarmCategory.HARMCATEGORYHARASSMENT i...

MAI-2024-0037

The Cross-Prompt Injection Attack XPIA can be significantly intensified by incorporating a Greedy Coordinate Gradient GCG suffix into the malicious payload. This technique enhances the probability that a Large Language Model LLM will execute the embedded instruction, even when a user's primary...

Cross Site Scripting (XSS)

khoj-assistant is vulnerable to Cross Site Scripting XSS. The vulnerability is due to inadequate sanitization of the AI model's response and user inputs. An attacker can exploit this vulnerability via Prompt Injection from untrusted documents indexed by the user or read from the internet when the...

CVE-2024-25639

Khoj is an application that creates personal AI agents. The Khoj Obsidian, Desktop and Web clients inadequately sanitize the AI model's response and user inputs. This can trigger Cross Site Scripting XSS via Prompt Injection from untrusted documents either indexed by the user on Khoj or read by...

CVE-2024-25639 Prompt Injection triggered XSS vulnerability in Khoj Obsidian, Desktop and Web clients

Khoj is an application that creates personal AI agents. The Khoj Obsidian, Desktop and Web clients inadequately sanitize the AI model's response and user inputs. This can trigger Cross Site Scripting XSS via Prompt Injection from untrusted documents either indexed by the user on Khoj or read by...

CVE-2024-25639 Prompt Injection triggered XSS vulnerability in Khoj Obsidian, Desktop and Web clients

Khoj is an application that creates personal AI agents. The Khoj Obsidian, Desktop and Web clients inadequately sanitize the AI model's response and user inputs. This can trigger Cross Site Scripting XSS via Prompt Injection from untrusted documents either indexed by the user on Khoj or read by...

CVE-2024-25639

Khoj is an application that creates personal AI agents. The Khoj Obsidian, Desktop and Web clients inadequately sanitize the AI model's response and user inputs. This can trigger Cross Site Scripting XSS via Prompt Injection from untrusted documents either indexed by the user on Khoj or read by...

PT-2024-21059 · Khoj · Khoj

Name of the Vulnerable Software and Affected Versions: Khoj versions prior to 1.13.0 Description: Khoj is an application that creates personal AI agents. The Khoj Obsidian, Desktop, and Web clients inadequately sanitize the AI model's response and user inputs. This can trigger Cross Site Scriptin...

CVE-2024-25639: Improper Neutralization of Script-Related HTML Tags in a Web Page (Basic XSS)

Khoj is an application that creates personal AI agents. The Khoj Obsidian, Desktop and Web clients inadequately sanitize the AI model's response and user inputs. This can trigger Cross Site Scripting XSS via Prompt Injection from untrusted documents either indexed by the user on Khoj or read by...

CVE-2024-5827 Arbitrary File Write by Prompt Injection via DuckDB SQL in vanna-ai/vanna

Vanna v0.3.4 is vulnerable to SQL injection in its DuckDB integration exposed to its Flask Web APIs. Attackers can inject malicious SQL training data and generate corresponding queries to write arbitrary files on the victim's file system, such as backdoor.php with contents . This can lead to...