Last Week in Cyber: OpenAI's GPT‑5.5‑Cyber, Daybreak, Microsoft's multi-model agentic security

Captain Patches Hook joins David to discuss the latest developments across cybersecurity: Federal officials considering slashing the patch deadline for U.S. agencies from two weeks to three days OpenAI’s GPT-5.5-Cyber, a new permissive model that removes safety restrictions for verified defenders OpenAI's Daybreak frontier AI platform, a direct response to Anthropic's Mythos Microsoft’s new MDASH (Multi-Model Agentic Scanning Harness) Verizon DBIR 2026 report, which shows vulnerability exploitation overtaking credential theft as the top breach vector