Anthropic has created a method to defend AI models against “jailbreaks” — unauthorized workarounds to get an AI model to do things it was trained not to do, like providing instructions for building chemical weapons. Called Constitutional Classifiers, the system was 95 percent effective in identifying and preventing jailbreaks of Anthropic’s Claude 3.5 Sonnet in a test environment. In an effort to drum up real-world red-teaming, the company offered cash prizes of up to $15,000 to anyone who could jailbreak its Sonnet AI model. After some 3,000 hours of attempts by 185 participants, none claimed an award. Now the company is offering additional incentives. Continue reading Anthropic Will Award Cash for Jailbreaking AI Defense System

The Library of Congress and U.S. Copyright Office just passed exemptions to the Digital Millennium Copyright Act (DMCA) that legalizes the so-called right to repair. Although the DMCA was created to prevent copyright piracy, it also resulted in a host of problematic side effects. Because devices such as smartphones come loaded with digital rights management (DRM) software, users infringed copyright laws if they attempted to repair such devices. With the new exemptions, users are now free to do so. Continue reading Library of Congress, Copyright Office Unlock Gadget Repair