ASL-3 | Seis News

Tag: ASL-3

Anthropic Upgrades Claude Opus 4 with Cutting-Edge AI Safety Level 3 Safeguards

Posted by Kris Stewart | May 27, 2025 | AI |

Anthropic has activated its strictest safety measures yet for Claude Opus 4, implementing AI Safety Level 3 protocols to prevent misuse in chemical, biological, radiological, and nuclear weapon development. The new safeguards target CBRN risks while maintaining a balance between safety and progress.

Anthropic Teams Up with HackerOne to Stress-Test Its AI Safeguards in New Bug Bounty Program

Posted by Kris Stewart | May 19, 2025 | AI |

A new $25K bug bounty program is testing unreleased AI safety classifiers to identify universal jailbreaks, focusing on CBRN-related risks as part of meeting the ASL-3 Deployment Standard. The initiative invites researchers to stress-test Claude 3.7 Sonnet’s safeguards, with rewards for verified vulnerabilities found before the program ends in May 2025.