OpenAI and Anthropic brief Congress on cyber-capable AI models.

This briefing summarizes recent cybersecurity and related national-security developments: researchers disclosed a high-severity Linux kernel flaw (CVE-2026-31431, “Copy Fail”) with public proof-of-concept enabling local privilege escalation; the pro-Iranian group Handala published personal data for ~2,300 US Marines; OpenAI and Anthropic briefed Congress on cyber-capable AI models; and the report notes legislative and supply-chain impacts (FISA Section 702 extension, geofencing warrant guidance, and disrupted circuit-board material supply).