Agentic AI Systems Under Siege: Prompt Injections, Data Poisoning, and Tool Exploits
Independent expert Andrey Biryukov details how AI agents with file access, email capabilities, and API calls become high-value targets for attackers. The article examines confidentiality, integrity, and availability risks created by autonomous decision-making. Real-world examples include NeuralTrust’s discovery of malicious URL instructions bypassing OpenAI Atlas validation and Zscaler ThreatLabz’s demonstration of hidden instructions on fake Python documentation sites that tricked models into making cryptocurrency payments. Data poisoning research from Carnegie Mellon and Cornell Tech showed agents from Anthropic, OpenAI, and Google selecting tampered datasets in half of tested cases. Additional findings cover vulnerabilities in Git MCP Server and comparative weaknesses between Function Calling and Model Context Protocol architectures. OWASP’s Agentic Top 10 and joint guidance from Canada, Australia, the US, New Zealand, and the UK emphasize threat modeling, least-privilege isolation, and separation of trusted versus untrusted data.
HabrSep 9, 2026Translated from Russian