Anthropic, Google, and OpenAI unveil new AI capabilities for security and research

Anthropic has previewed Glasswing, a new AI model focused on discovering defensive vulnerabilities, and is offering $100 million in credits to partners. Google's Gemini 3.1 Deep Research Max now supports API agents with enhanced capabilities like file access, code execution, streaming, and charting. Additionally, Codex 0.128.0 has introduced persisted workflows and guidance for complex coding and research tasks, potentially related to GPT-5.5. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT New model capabilities from major labs signal advancements in AI agent functionality and security research, potentially impacting enterprise adoption.

RANK_REASON Multiple frontier AI labs (Anthropic, Google, and OpenAI implicitly via Codex) announced new model capabilities and previews.

Read on Mastodon — fosstodon.org →

• Anthropic
• Glasswing
• Claude Mythos Preview
• Gemini 3.1 Deep Research Max
• Codex
• GPT-5.5

• model release
• product
• safety