AI Governance Institute
← All news

Agentic AI Threats

Agentic AI threats are security and safety risks that arise when AI systems act autonomously, such as by browsing the web, running code, calling software tools, or taking actions on behalf of users without step-by-step human approval. The core concern is that an attacker or a simple error can cause an agent to misuse its own permissions, for example by being tricked through hidden instructions planted in a document or web page into leaking data or making unauthorized purchases. For compliance and risk teams, these threats matter because they expand the potential for an AI system to cause real operational, financial, and legal harm well beyond the incorrect text a chatbot might produce.

1 item