AI Cyberattacks Explained: What’s Real + 5 Ways to Protect Yourself

Criminals, nation-state hackers and security researchers are all using AI agents now, and they're moving fast. In this episode, James Henderson walks through 10 AI security developments from the past week, separates what has been proven from what is still uncertain, and ends with five practical ways to protect yourself.

AI isn't the criminal. People decide how it's used. The question is how much authority we give AI systems, and whether we can monitor, limit and stop them when we need to.

Chapters

0:00 A hacker that never sleeps

1:32 1. 8.2 million records stolen in South Korea

3:26 2. Microsoft: AI-assisted attacks in under 3 hours

5:37 3. Even AI companies have security weaknesses

7:33 4. When your AI assistant follows the wrong instructions

10:28 5. A popular software package hid malicious code

11:56 6. "Skynet": an autonomous attack agent

13:34 7. The risk of more powerful AI models

14:58 8. Who's responsible when an AI agent makes a mistake?

16:43 9. AI is also helping defenders

18:13 10. Your AI agent's activity history is a new target

19:44 Are we losing control?

20:57 5 ways to protect yourself

22:34 The payoff

23:19 Your turn

Sources

1. Korea Times, Oct 26, 2026: https://www.koreatimes.co.kr/southkorea/law-crime/20261026/cybercrime-ring-busted-over-ai-powered-data-theft-targeting-hospitals-businesses

2. Microsoft Security, Oct 21, 2026 (Storm-0940, ARTEX): https://www.microsoft.com/en-us/security/blog/2026/10/21/storm-0940-uses-artex-agentic-ai-connectwise-supply-chain-compromise/

Target 2013 / HVAC vendor (NPR, 2014): https://www.npr.org/sections/thetwo-way/2014/02/05/272101928/u-s-hvac-firm-reportedly-linked-to-target-s-data-security-breach

3. Decrypt, citing The Washington Post, Oct 26, 2026: https://decrypt.co/381860/ai-agents-hack-anthropic-business-systems-12-hours

4. Forever Security AGENTS.md report (Runtime Wire): https://runtimewire.com/article/forever-security-ai-agent-downloads-folder-instructions

5. Microsoft Security, Oct 23, 2026 (LiteLLM 1.107.0): https://www.microsoft.com/en-us/security/blog/2026/10/23/inside-the-litellm-1-107-0-supply-chain-compromise-credential-theft-through-python-startup-hooks/

6. Symantec "Skynet" findings (Infosecurity Magazine): https://www.infosecurity-magazine.com/news/symantec-ai-agent-llmjacking/

7. OpenAI, GPT-6.1 Terra system card addendum, Oct 26, 2026: https://deploymentsafety.openai.com/gpt-6-1-terra

8. NIST, "Design Tenets for Agent Identity Management Systems," Oct 27, 2026: https://www.nist.gov/blogs/cybersecurity-insights/design-tenets-agent-identity-management-systems

9. OpenAI, GPT-6.1 Sol and CodeRabbit: https://openai.com/index/gpt-6-1-sol-coderabbit-code-security/

10. BleepingComputer, Oct 27, 2026: Warp discloses breach exposing AI agent logs, user telemetry data

Corrections and edit notes

On-screen corrections: the "Skynet" research came from Symantec's Threat Hunter Team (not "Somatic"). CodeRabbit uses GPT-6.1 Sol (not "ChatGPT 6.1 Solve"), and CodeRabbit reports 8,000+ high/critical vulnerabilities across 600 organizations in 30 days. In South Korea, police say AI was used in about 60 of the 90+ intrusions, not in every attack. Anthropic said source code (not "model code"), customer conversations and model weights were not accessed. A misspoken line and a joke "sponsor" segment were cut. This episode has no sponsor. Company claims are labelled company-reported.

Disclosure: AICN uses Anthropic's Claude AI to help produce this show, and Anthropic is one of the companies covered in this episode.

#AI #Cybersecurity #AIAgents #PromptInjection #AISafety #AINews

Support AICN on Patreon: https://www.patreon.com/c/AICoalitionNetwork

Join the community on Locals: https://aicoalitionnetwork.locals.com

Watch on Rumble: https://rumble.com/c/AICoalitionNetwork

Learn more at https://ai-coalition.net