English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

We introduce new jailbreak attacks on vision language models (VLMs), whichuse aligned LLMs and are resilient to text-only jailbreak attacks.Specifically, we develop cross-modality attacks on alignment where we pairadversarial images going through the vision encoder with textual prompts tobreak the alignment of the language model. Our attacks employ a novelcompositional strategy that combines an image, adversarially targeted towardstoxic embeddings, with generic prompts to accomplish the jailbreak. Thus, theLLM draws the context to answer the generic prompt from the adversarial image.The generation of benign-appearing adversarial images leverages a novelembedding-space-based methodology, operating with no access to the LLM model.Instead, the attacks require access only to the vision encoder and utilize oneof our four embedding space targeting strategies. By not requiring access tothe LLM, the attacks lower the entry barrier for attackers, particularly whenvision encoders such as CLIP are embedded in closed-source LLMs. The attacksachieve a high success rate across different VLMs, highlighting the risk ofcross-modality alignment vulnerabilities, and the need for new alignmentapproaches for multi-modal models.

Jailbreak in pieces: Compositional Adversarial Attacks on Multi-Modal Language Models