Anthropic

A weekly roundup of the biggest developments shaping the global AI landscape

GPT-5.4-Cyber is aimed at vetted defenders working on malware analysis, vulnerability research, and reverse engineering under OpenAI’s Trusted Access for Cyber framework.

The restricted program gives select technology companies and software maintainers early access to Anthropic’s unreleased Claude Mythos Preview model for defensive cyber work.

The slip-up sent developers hunting for clues and sharpened concerns about security and disclosure practices

Based on interviews through December 2025 with over 80,000 users across 159 countries, the findings capture a nuanced snapshot of how people perceive AI—the good, the underwhelming, and its potential.