AI Incident Detection

AI incident detection refers to the methods and tools organizations use to spot when an AI system is behaving harmfully, failing, or being misused, such as a chatbot leaking customer data, a model producing discriminatory decisions, or an attacker manipulating an AI tool. For compliance and risk teams, it is the operational side of AI accountability: without monitoring that flags these events, a company may not learn of a problem until regulators, customers, or the press do. Strong detection programs also support legal obligations that require timely reporting of serious AI-related incidents under frameworks such as the EU AI Act.

1 item

Corporate PolicyGlobal2026-10-28

Cisco's Project Quarry Exposes the Detection Gap in Enterprise Agent Monitoring

Cisco has open-sourced Project Quarry, a behavioral detection engine that watches AI agent sessions for multi-step attack patterns. Existing security tools miss these attacks because each individual step looks harmless; only the sequence reveals malicious intent. The release makes behavioral monitoring for AI agents a concrete, achievable control rather than an aspirational one.

Cisco agentic AI AI security behavioral monitoring AI incident detection open source agent oversight