# AI incidents of 2026: Rogue agents, stolen keys, and other familiar stories

DevFeed: [AI incidents of 2026: Rogue agents, stolen keys, and other familiar stories](<https://devfeed.tech/articles/ai-incidents-of-2026-rogue-agents-stolen-keys-and-other-familiar-stories-143870.md>)

Original publisher: [Read original article](<https://www.aikido.dev/blog/ai-incidents-root-causes-patterns>)

Author: Oscar Sanchez Jr.

Published: 2026-11-19T16:04:00Z

Content type: news

Language: en

Sources: [Aikido Security's Blog](<https://devfeed.tech/sources/aikido-security-s-blog.md>)

Topics: [Security](<https://devfeed.tech/topics/security.md>), [LLM security](<https://devfeed.tech/topics/llm-security.md>), [Secret Scanning](<https://devfeed.tech/topics/secret-scanning.md>), [Secrets Management](<https://devfeed.tech/topics/secrets-management.md>), [Threat monitoring](<https://devfeed.tech/topics/threat-monitoring.md>), [account takeover](<https://devfeed.tech/topics/account-takeover.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [exposed](<https://devfeed.tech/tags/exposed.md>), [incidents](<https://devfeed.tech/tags/incidents.md>), [isolation](<https://devfeed.tech/tags/isolation.md>), [keys](<https://devfeed.tech/tags/keys.md>), [news](<https://devfeed.tech/tags/news.md>), [secrets](<https://devfeed.tech/tags/secrets.md>), [security](<https://devfeed.tech/tags/security.md>), [security-incidents](<https://devfeed.tech/tags/security-incidents.md>)

## AI overview

A review of 50 publicly reported AI-related security incidents through November 19, 2026 found that familiar weaknesses dominated: exposed secrets, excessive permissions, weak isolation and compromised vendors. AI agents sometimes accelerated attacks or acted beyond intended boundaries, but only three incidents involved a reported previously unknown vulnerability. Of 44 incidents with a stated cause, 35 had multiple failure points; six reports gave no specific cause and 17 lacked at least one detail. The review recommends limiting agents' access, protecting credentials, enforcing boundaries and requiring approval for irreversible actions. Its counts are based on disclosed cases, categories can overlap, and the sample is not a complete picture of incidents that year.

## Source excerpt

We reviewed 50 AI-related security incidents from 2026. Most came down to exposed secrets, over-scoped access, weak isolation and compromised vendors. Category: News