Particle.news
Download on the App Store

IMF Warns of AI Cyber Threat to Finance as Access to Anthropic’s Mythos Stays Locked Down

Officials urge rapid hardening to keep pace with models that chain software flaws into working hacks.

Overview

  • - The IMF, which issued a warning Thursday, says fast-moving AI tools could turn a single breach into a market-shaking shock and calls for stronger resilience, supervision, and cross-border coordination.
  • - Australia’s securities regulator told financial firms to act now to shore up basic cyber hygiene, warning that new AI can uncover weaknesses faster than companies can patch them.
  • - OpenAI introduced a limited preview of GPT-5.5-Cyber for vetted security teams to speed tasks like finding bugs, checking patches, and analyzing malware.
  • - Anthropic is keeping Claude Mythos in a restricted program for about 40 partners under Project Glasswing, leaving many banks and regulators outside the U.S. seeking access to harden critical software.
  • - Cyber researchers told CNBC they can match many Mythos-style results by linking existing models and tools in coordinated workflows, a sign these capabilities may spread quickly and raise pressure to fix vulnerable code.