← All predictionsCyber Predictions

No. 20Said

UnderwayOpen models

Open models at frontier cyber level let attackers work without lab monitoring; only skill and compute will constrain them.

X post, June 13, 2026 

First, it would allow this level of malicious cyber activity without needing to hide from Anthropic / OpenAI. The only constraint would be attacker skill and compute.

Daniel Miessler, 13 June 2026

First evidence: 3 months later

Jun 2026 → Sep 2026
  1. Anthropic: open-weight GLM-5.3 builds end-to-end exploits like Mythos Preview, outside any lab's safeguardsanthropic.com
  2. CrowdStrike: ARTEX intrusions against South Korean finance ran on open DeepSeek and GLM modelscrowdstrike.com

He said it again

Where this call sits in the record

Every call, 2016 to now

When he said itFirst outside evidenceNo evidence yetHas happened Underway Not yetLinear time: equal years, equal length