**Anthropic: Model leaks, Pentagon lawsuit, ethics/safety crises** [developing]
Key Questions
What is Claude Mythos Preview?
Claude Mythos Preview is Anthropic's advanced AI model that excels at identifying thousands of zero-day vulnerabilities in major operating systems and web browsers. It has demonstrated a 72% success rate in chaining exploits. The model is confirmed too dangerous for public release due to risks of enabling cyberattacks.
Why is Anthropic limiting the rollout of Mythos?
Anthropic fears hackers could use Mythos for faster cyberattacks, so it is restricting access initially to defenders at Microsoft, Amazon, Google, and Apple. The government has been briefed on its capabilities. A full public release has been deemed too risky.
What is Project Glasswing?
Project Glasswing is an Anthropic initiative where the AI autonomously hunts thousands of zero-day vulnerabilities in OS and browsers. It achieved 72% chaining success and is part of efforts to address critical bugs, some unpatched for decades. It supports a limited rollout to key defenders first.
What vulnerabilities has Mythos identified?
Mythos has found thousands of zero-day vulnerabilities in every major operating system and web browser, sparking a race to fix critical bugs. Some vulnerabilities remained unpatched for decades. This has prompted urgent patching efforts.
Who gets early access to Mythos?
Limited rollout is prioritized for defenders at Microsoft, Amazon, Google, and Apple. The government has been briefed. This gating aims to bolster defenses before wider availability.
What prior leaks are associated with Anthropic?
Apparent leaks revealed details of the unreleased Mythos product late last month. Anthropic also leaked 500,000 lines of secret code, with contents raising further concerns. These incidents highlight ongoing ethics and safety crises.
What is the Pentagon lawsuit against Anthropic?
A $5B lawsuit from the Pentagon continues amid Anthropic's model developments. It relates to broader ethics and safety issues. Details tie into prior controversies like OpenClaw ban.
What is the OpenClaw ban in Claude subscriptions?
OpenClaw was banned in Claude subscriptions, leading to an 8.9% usage surge in GPT-5.4. This reflects safety measures amid agentic tool concerns. It coincides with Mythos-related restrictions.
Mythos/Claude Mythos Preview confirmed too dangerous for public release; Project Glasswing autonomously hunts thousands zero-days in OS/browsers (72% chaining success); limited rollout to MS/Amazon/Google/Apple defenders first, gov briefed; prior leaks, OpenClaw ban, $5B Pentagon suit continue.