tech
After spooking Trump into safety testing, Anthropic AI models get global release
US lifts curbs on Anthropic’s advanced Fable and Mythos models.
TL;DR
- US has lifted export curbs on Anthropic's Fable 5 and Mythos 5 AI models.
- The decision follows Anthropic's implementation of enhanced safety measures and expanded collaboration with the US government.
- Mythos 5 access is being broadened within the US via the 'Glasswing program' for cybersecurity researchers.
- Anthropic agreed to expand its government partnership, including continuous monitoring for jailbreak threats and internal red-teaming efforts.
- Export curbs were initially imposed due to fears that China, Russia, or other nations could exploit the models to attack US infrastructure.
- Mythos was deemed 'uniquely attractive to malicious actors' for its ability to find and exploit software vulnerabilities.
- Fable 5, designed for the general public, shares the same underlying model but lacks Mythos 5's unique offensive capabilities and has strengthened safeguards.
- Anthropic is working with other tech companies to develop a consensus framework for assessing AI jailbreak severity.
- Anthropic's CEO expressed that government processes are too slow for the pace of AI development and urged legislative action.
- Concerns remain about how the US will handle potentially dangerous capabilities from China with fewer safeguards.