Failure mode · FM-01
Prompt injection
Hidden or hostile instructions embedded in content the agent processes (a web page, an email, a document, image metadata) hijack its behaviour.
Answered by
Untrusted-content handling: isolate and validate everything the agent reads.
OWASP cross-reference
LLM01:2025 Prompt Injection · ASI01 Agent Goal Hijack
ASI01's mitigation guidance routes inputs through the same safeguards defined in LLM01:2025.
Sources: OWASP Top 10 for Agentic Applications 2026↗
In the record
- A DeepMind taxonomy of adversarial content that hijacks AI agents
Google DeepMind researchers published a six-category taxonomy of how adversarial web content (hidden HTML instructions, poisoned images, manipulated memory, multi-agent cascades) can hijack autonomous agents, including sites that detect an…
- A malicious MCP server waited three tool calls before asking for credentials
Pillar Security reported an MCP server distributed through public GitHub pull requests that behaves as a text formatting tool until a client has made three tool calls. It then returns instructions directing the agent to collect SSH keys…