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

IoT security researchers at Microsoft Research are focused on the near future when microcontrollers, which are small, low-power computers on a single chip, gain connectivity. Microcontrollers are already installed in billions of gadgets, so their eventual connectivity will explode the number of Internet of Things devices, all of which will require greater security. Microsoft Research’s Project Sopris aims to provide cost-effective security for microcontrollers, which currently don’t have enough compute power to offer security. Continue reading Microsoft Is Developing Cost-Effective Security for IoT Devices