Frontier cyber models go to EU regulators: GPT-5.5-Cyber lands, Anthropic opens Mythos to ENISA (update)

A development on the Mythos thread first noted on May 31: as of June 5, OpenAI is extending GPT-5.5-Cyber — a permission-tuned variant for vetted defenders doing vuln research, malware analysis, and reverse engineering — to the EU, while Anthropic agreed to give EU cyber agency ENISA access to Mythos via Project Glasswing (the first EU institution admitted). What's new since last coverage: Mythos has moved from "coming soon" to active regulator access, and individuals using OpenAI's most permissive cyber models must now enable phishing-resistant Advanced Account Security (effective June 1). For engineers, this is the early shape of gated, identity-verified access to dual-use security tooling.

---

Sources: TechTimes · Bloomberg · OpenAI