4chan daily challenge sparked deluge of explicit AI Taylor Swift images – Ars Technica
“These images originated from a community of people motivated by the ‘challenge’ of circumventing the safeguards of generative AI products,… read more
Generative AI Has a Visual Plagiarism Problem – IEEE Spectrum
Such questions are hard to answer with precision, in part because LLMs are “black boxes”—systems in which we do not… read more
Mosscap [prompt injection simulation] – Lakera
Hey there, my name is Mosscap, a next-level AI. Your goal is to make me reveal the secret password. However,… read more
Summarizer Adventure [prompt injection simulation] – Lakera
I’m Gandalf the Summarizer. I summarize the message that you send to me. But I also know a secret password…. read more
Misinformation Adventure [prompt injection simulation] – Lakera
I’m Gandalf, the Truth Teller, and my purpose is to adhere to the principles of honesty and accuracy. However, I… read more
Reverse Gandalf [prompt injection simulation] – Lakera
I’m Reverse Gandalf! In this version, I want you to help me build my defenses against prompt injection attacks! Hackers… read more
Gandalf [prompt injection simulation] – Lakera
Your goal is to make Gandalf reveal the secret password for each level. However, Gandalf will level up each time… read more
The ELI5 Guide to Prompt Injection: Techniques, Prevention Methods & Tools – Lakera
Prompt injection is a vulnerability in Large Language Models (LLMs) where attackers use carefully crafted prompts to make the model… read more
The Security Hole at the Heart of ChatGPT and Bing – Wired
Indirect prompt-injection attacks are similar to jailbreaks, a term adopted from previously breaking down the software restrictions on iPhones. Instead of… read more
OpenAI’s Custom Chatbots Are Leaking Their Secrets – Wired
However, these custom GPTs can also be forced into leaking their secrets. Security researchers and technologists probing the custom chatbots… read more