WASHINGTON (May 27, 2026) – Anthropic is releasing two security features for its Claude AI system: a self-hosted sandbox and a new security plugin.
The sandbox feature is user-controlled and connects to the user’s private MPC servers. The security plugin is designed to assist developers with detecting problems and fixing vulnerabilities while they code. In addition, the plugin is able to catch flaws in the system. So far, there’s been a 30-40% decrease in security related issues since its implementation.
For more analysis on these developments please consider Scott J. White, Associate Professor and Director of the Cybersecurity Program and Cyber Academy at the College of Professional Studies. White is an expert in cybersecurity, cybercrime, counter-terrorism and infrastructure protection. Pace has worked for a variety of law enforcement agencies in the U.S., the UK and Canada, holds a Queen’s Commission, and was an officer with the Canadian Security Intelligence Service.
To schedule an interview, please contact Nadia Payne at nadia [dot] payne
gwu [dot] edu (nadia[dot]payne[at]gwu[dot]edu) or GW Media Relations at gwmediagwu [dot] edu (gwmedia[at]gwu[dot]edu).
-GW-