Claude Mythos Zero-Days, OpenAI's 2026 Election Defense & GPT-5.5 Agents

On this AI RADAR, we explore major shifts in AI cybersecurity and agentic development. **Anthropic** unveils Project Glasswing and **Claude Mythos Preview**, an unreleased model that autonomously found 10,000+ zero-day vulnerabilities. **OpenAI** introduces 2026 election safeguards, actively defending voting infrastructure from state-sponsored attacks. Also, developer tool **Warp** partners with OpenAI to deploy autonomous coding agents using the highly efficient **GPT-5.5**. Plus, new research on the **ExploitGym** cyber-attack benchmark and an Anthropic study revealing the gap in coding agent adoption among social scientists. Sources: • Project Glasswing: https://www.anthropic.com/news/project-glasswing • Election Safeguards: https://openai.com/index/election-safeguards-2026/ • Warp & GPT-5.5: https://openai.com/index/warps-big-bet-on-building-open-source-with-gpt-5-5 • Social Science Survey: https://www.anthropic.com/research/coding-agents-social-sciences • ExploitGym: https://arxiv.org/abs/2605.11086