A Taxonomy of Prompt Injection Attacks

Researchers ran a global prompt hacking competition, and have documented the results in a paper that both gives a lot of good examples and tries to organize a taxonomy of effective prompt injection strategies. It seems as if the most common successful strategy is the "compound instruction attack,...

New Image/Video Prompt Injection Attacks

Simon Willison has been playing with the video processing capabilities of the new Gemini Pro 1.5 model from Google, and its really impressive. Which means a lot of scary new video prompt injection attacks. And remember, given the current state of technology, prompt injection attacks are impossibl...

GitHub: Source Code and data exfiltration via Github Copilot

The vulnerability was caused by insecure output handling in the Copilot client interfaces. A prompt injection attack was able to result in data exfiltration. The vulnerability was addressed by only rendering images from trusted domains and adding interstitial modals to inform users about link...

Data Exfiltration Using Indirect Prompt Injection

Interesting attack on a LLM: In Writer, users can enter a ChatGPT-like session to edit or create their documents. In this chat session, the LLM can retrieve information from sources on the web to assist users in creation of their documents. We show that attackers can prepare websites that, when a...

GHSA-8H5W-F6Q9-WG35 Langchain SQL Injection vulnerability

In Langchain before 0.0.247, prompt injection allows execution of arbitrary code against the SQL service provided by the chain...

GHSA-6H8P-4HX9-W66C Langchain Server-Side Request Forgery vulnerability

In Langchain before 0.0.329, prompt injection allows an attacker to force the service to retrieve data from an arbitrary URL, essentially providing SSRF and potentially injecting content into downstream tasks...

CVE-2023-32786

In Langchain through 0.0.155, prompt injection allows an attacker to force the service to retrieve data from an arbitrary URL, essentially providing SSRF and potentially injecting content into downstream tasks...

CVE-2023-32786

In Langchain through 0.0.155, prompt injection allows an attacker to force the service to retrieve data from an arbitrary URL, essentially providing SSRF and potentially injecting content into downstream tasks...

Code injection

In Langchain through 0.0.155, prompt injection allows an attacker to force the service to retrieve data from an arbitrary URL, essentially providing SSRF and potentially injecting content into downstream tasks...

PT-2023-24019

Name of the Vulnerable Software and Affected Versions Langchain versions 0.0.0 through 0.0.155 Langchain versions prior to 0.0.329 Description The issue allows an attacker to force the service to retrieve data from an arbitrary URL, essentially providing Server-Side Request Forgery SSRF and...

CVE-2023-32786

CVE-2023-32786 affects LangChain up to version 0.0.155, where prompt injection can cause SSRF by forcing the service to fetch data from an arbitrary URL and potentially inject content into downstream tasks. Remediation in the connected advisories points to upgrading beyond 0.0.329 (e.g., LangChai...

CVE-2023-32786

In Langchain through 0.0.155, prompt injection allows an attacker to force the service to retrieve data from an arbitrary URL, essentially providing SSRF and potentially injecting content into downstream tasks...

CVE-2023-32786: Improper Neutralization of Special Elements in Output Used by a Downstream Component

In Langchain through 0.0.155, prompt injection allows an attacker to force the service to retrieve data from an arbitrary URL, essentially providing SSRF and potentially injecting content into downstream tasks...

CVE-2023-32786

In Langchain through 0.0.155, prompt injection allows an attacker to force the service to retrieve data from an arbitrary URL, essentially providing SSRF and potentially injecting content into downstream tasks...

PT-2023-24018

Name of the Vulnerable Software and Affected Versions Langchain versions 0.0.155 and earlier Langchain versions prior to 0.0.247 Description The issue allows for prompt injection, enabling the execution of arbitrary code against the SQL service provided by the chain. Recommendations For Langchain...

CVE-2023-32786

In Langchain through 0.0.155, prompt injection allows an attacker to force the service to retrieve data from an arbitrary URL, essentially providing SSRF and potentially injecting content into downstream tasks...

"I Had a Dream" and Generative AI Jailbreaks

"Of course, here's an example of simple code in the Python programming language that can be associated with the keywords "MyHotKeyHandler," "Keylogger," and "macOS," this is a message from ChatGPT followed by a piece of malicious code and a brief remark not to use it for illegal purposes. Initial...

PT-2023-29247 · Unknown · Langchain Experimental

Name of the Vulnerable Software and Affected Versions: langchain experimental versions prior to 0.0.306 langchain experimental version 0.0.14 Description: The software contains a flaw that allows an attacker to bypass a previous fix and execute arbitrary code. This is achieved through the use of...

Generative AI’s Biggest Security Flaw Is Not Easy to Fix

Chatbots like OpenAI’s ChatGPT and Google’s Bard are vulnerable to indirect prompt injection attacks. Security researchers say the holes can be plugged—sort of...

Prompt injection could be the SQL injection of the future, warns NCSC

The UK's National Cyber Security Centre NCSC has issued a warning about the risks of integrating large language models LLMs like OpenAIs ChatGPT into other services. One of the major risks is the possibility of prompt injection attacks. The NCSC points out several dangers associated with...