ARTFEED — Contemporary Art Intelligence

Anthropic's Mythos AI Sparks Cybersecurity Fears and China Debate

ai-technology · 2026-04-22

Anthropic's new AI model, Claude Mythos Preview, announced on April 7, has raised global cybersecurity concerns due to its ability to identify and exploit vulnerabilities. Instead of a public release, it was given to a consortium including Cisco, JPMorgan Chase, and Nvidia under Project Glasswing to secure critical software. The model's potential sharing with China is debated, as Chinese open-source models lag behind US closed-source ones in cyber capabilities but are improving rapidly, according to Beijing-based Concordia AI.

Key facts

  • Anthropic announced Claude Mythos Preview on April 7.
  • Mythos can identify and exploit cybersecurity vulnerabilities.
  • Released to a consortium of US companies: Cisco, JPMorgan Chase, Nvidia.
  • Part of Project Glasswing to secure critical software.
  • Debate over whether capabilities should be shared with China.
  • Chinese open-source models lag behind US closed-source models in cyber capabilities.
  • Chinese models are improving rapidly, per Concordia AI.
  • Concordia AI is a Beijing-based consultancy.

Entities

Institutions

  • Anthropic
  • Cisco
  • JPMorgan Chase
  • Nvidia
  • Project Glasswing
  • Concordia AI

Locations

  • Beijing
  • China
  • United States

Sources