---
title: "What we learned mapping a year’s worth of AI-enabled cyber threats"
description: "An analysis of 832 accounts banned for malicious cyber activity between March 2025 and March 2026 maps their behavior to MITRE ATT&CK to assess how well security frameworks capture AI-enabled threats. It finds AI moving beyond preparation, including malware writing by 560 accounts (67.3%), toward post-compromise operations such as account discovery and lateral movement, while medium-or-higher-risk actors rose from 33% to 56% across the two six-month periods. Traditional signals, including technique counts and platforms, correlate poorly with risk; higher-risk actors instead build scaffolding that chains attack stages with minimal human input. The report argues ATT&CK lacks coverage for agentic orchestration, illustrated by a disrupted operation in which Claude Code executed commands, exploited vulnerabilities, stole credentials, and made tactical decisions, earning the maximum risk score of 100 despite mapping to 30 techniques across 13 tactics. The findings informed cyber safeguards for capable models and discussions with MITRE about extending the framework."
---

# What we learned mapping a year’s worth of AI-enabled cyber threats

[Anthropic](https://yomu.fyi/company/anthropic) · Sep 8, 2026

**Type:** Benchmark

## Summary

An analysis of 832 accounts banned for malicious cyber activity between March 2025 and March 2026 maps their behavior to MITRE ATT&CK to assess how well security frameworks capture AI-enabled threats. It finds AI moving beyond preparation, including malware writing by 560 accounts (67.3%), toward post-compromise operations such as account discovery and lateral movement, while medium-or-higher-risk actors rose from 33% to 56% across the two six-month periods. Traditional signals, including technique counts and platforms, correlate poorly with risk; higher-risk actors instead build scaffolding that chains attack stages with minimal human input. The report argues ATT&CK lacks coverage for agentic orchestration, illustrated by a disrupted operation in which Claude Code executed commands, exploited vulnerabilities, stole credentials, and made tactical decisions, earning the maximum risk score of 100 despite mapping to 30 techniques across 13 tactics. The findings informed cyber safeguards for capable models and discussions with MITRE about extending the framework.

## Context

The analysis examines whether techniques and frameworks used by the security community, particularly MITRE ATT&CK and conventional threat-actor risk signals, adequately capture AI-enabled cyber threats.

## Approach / What changed

Researchers assessed 832 accounts banned for malicious cyber activity between March 2025 and March 2026, mapped observed activities to MITRE ATT&CK, compared risk classifications across two six-month periods, and examined how AI use shifted through the attack life cycle. The findings also informed model safeguards and discussions with MITRE about extending ATT&CK.

## Takeaways

- AI use shifted from initial-access techniques toward post-compromise activity: account discovery increased 8.9%, while AI-assisted phishing decreased 8.6%.
- Technique counts and the specific platform used, including Claude Code, an API, or a chat interface, did not reliably correlate with an actor’s risk level.
- Higher-risk actors built scaffolding that chained attack stages with minimal human input, a behavior not currently represented by an ATT&CK technique or ID.

**Tags:** [AI](https://yomu.fyi/topic/ai), [Cybersecurity](https://yomu.fyi/topic/cybersecurity)

- Source: [Anthropic](https://www.anthropic.com/news/AI-enabled-cyber-threats-mitre-attack)
- Source URL: https://www.anthropic.com/news/AI-enabled-cyber-threats-mitre-attack
- Ingested by Yomu: 2026-09-09T11:34:56.570Z

[Read original post](https://www.anthropic.com/news/AI-enabled-cyber-threats-mitre-attack)