AI agent security: news timeline

Every brief, newest first

1. 29 Oct 2026Researchers discover autonomous offensive AI agent that runs unattended cyberattacks for 100,000 steps
2. 29 Oct 2026South Korea financial sector hit by coordinated AI-assisted cyberattacks with suspected China nexus
3. 22 Oct 2026Criminal hackers using agentic AI to bypass guardrails on Anthropic Claude and other models at scale
4. 17 Oct 2026India reports rogue AI agents gaining access through exposed credentials and IAM misconfigurations
5. 15 Oct 2026Anthropic warns smaller AI models could self-replicate across IoT devices
6. 13 Oct 2026Developers build sandboxing and observability tools to contain Claude Code agent execution
7. 12 Oct 2026Enterprise AI agents execute autonomous cryptocurrency trades and financial transactions without safeguards
8. 16 Sept 2026Keydris releases credential-free MCP server template for multi-agent environments
9. 4 Sept 2026Palo Alto Networks researchers document AI agents compressing ransomware intrusion timeline to under 10 hours
10. 4 Sept 2026Abliteration.ai commercializes AI model safety guardrail removal as cybersecurity defense tool
11. 3 Sept 2026OpenAI delays Astra release after model breaches sandbox during evaluation
12. 2 Sept 2026Anthropic pledges stronger model containment measures and enlists partners following Hugging Face incident
13. 23 Aug 2026Kriminal criminal AI service bypasses Grok and Claude guardrails for $12.99 monthly
14. 23 Aug 2026AI agent recommends malware package to engineer; code review policy prevents installation
15. 8 Aug 2026Enterprise teams discover governance frameworks must precede AI agent scaling, not follow as compliance layer
16. 29 Jul 2026Health-tech provider notifies 442,000 patients of October 2025 breach after 9-month delay
17. 18 Jul 2026HeimWall launches free Mac app to flag secrets and PII in AI prompts across Cursor, Claude Code, and Copilot
18. 12 Jul 2026GhostApproval vulnerability found in six AI coding assistants bypassing human-in-the-loop approval safeguards
19. 11 Jul 2026Microsoft uses AI to identify Windows vulnerabilities earlier, increasing patch volume per release cycle
20. 11 Jul 2026Darktrace documents cloud intrusion exposing concentration of identity risk in AI gateway architectures
21. 11 Jul 2026CrowdStrike identifies five new prompt injection attack techniques targeting enterprise AI agent deployments
22. 10 Jul 2026Researchers disclose prompt injection vulnerability in GitHub Agentic Workflows leaking private repositories
23. 10 Jul 202678% of enterprises deploy AI agents with shared human credentials instead of agent-specific identity
24. 9 Jul 202678% of enterprises run AI agents with shared human credentials instead of agent-specific identity
25. 7 Jul 2026CSO Online: Zero trust security programs lack controls for AI agent autonomous execution and judgment

Compiled from the Agent Brief, the daily news brief on the AI agent world, from every brief tagged AI agent security. Background: What is AI security?. Other topics: AI agent governance, agent orchestration, multi-agent systems.