Two leading US AI companies launched cybersecurity-capable frontier models in the weeks prior to 19 May 2026.
95% confidence
OtherNorth America
Omissions
The MEP did not name the two companies explicitly, leaving the reference ambiguous, though the launches of Anthropic Mythos and OpenAI Daybreak are the clear referents.
The claim uses the term 'models' broadly; OpenAI's Daybreak is more accurately described as an initiative or platform combining a model (GPT-5.5) with a dedicated security harness (Codex Security) rather than a single standalone model.
Sources
PrimaryAnthropic (Project Glasswing)On April 7, 2026, Anthropic announced Project Glasswing, a new initiative bringing together Amazon Web Services, Anthropic, Apple, and others to secure critical software, coinciding with the preview release of its frontier AI model Claude Mythos, whose advanced cybersecurity capabilities prompted a restricted rollout.
PrimaryOpenAI (Daybreak)OpenAI launched Daybreak, deploying AI for cyber defense using GPT-5.5 and Codex Security to identify threats, generate patches, and verify remediation across code and systems.
SecondaryTechCrunchAnthropic debuted a preview of its powerful new AI model Mythos on April 7, 2026, within a new cybersecurity initiative. The model represented a 'step change' in AI capabilities for cybersecurity.
SecondaryCNBCAnthropic limited the rollout of its Mythos AI model over cyberattack fears, granting access only to Microsoft, Amazon, Apple, CrowdStrike, Palo Alto Networks and roughly 40 other companies for defensive security purposes.
SecondaryCIO DiveOpenAI announced a cybersecurity initiative called Daybreak on May 11, 2026, described as the company's answer to Anthropic's Mythos model, combining GPT-5.5 with Codex Security.
SecondaryFuturum GroupOpenAI launched Daybreak on May 11, 2026, combining the GPT-5.5 model family with Codex Security as an agentic harness and partner ecosystem for application security.