Multimodal Agents

Multimodal agents are AI systems that can take in and act on several types of input at once, such as text, images, audio, and video, and then carry out multi-step tasks like filling in forms, navigating software, or making purchases on a user's behalf. Because they operate across screens and data sources with limited human review at each step, they can access, change, or send sensitive information in ways that are hard to predict or log. For enterprise governance teams, this raises practical questions about permissions, audit trails, data handling, and who is accountable when an agent makes an error.

1 item

ResearchGlobal2026-11-11

Visual Prompt Injection in Multimodal Agents Exposes a Tool-Call Control Gap

Researchers publishing in the ACM have demonstrated an adaptive visual prompt injection technique that manipulates AI agents processing untrusted images, websites, documents, or emails. The attack can cause agents to leak personal information or make unauthorized tool calls without requiring access to the model's internals. Enterprises deploying multimodal agents in document processing, email handling, or web browsing workflows face a control gap that existing text-based input filters do not address.

prompt injection multimodal agents agentic AI tool calls red teaming data privacy agent security