In a significant surge, the open-source AI agent “OpenClaw” saw its publicly exposed deployments jump from 1,000 to over 21,000 within a week, raising alarms over security risks associated with its deployment in corporate environments. This rapid adoption has highlighted the dangers of prompt injection, a critical vulnerability that can allow attackers to manipulate AI agents and exfiltrate data without detection. In response, Cloudflare has introduced the Moltworker framework, a cloud-based sandbox solution that enables organizations to evaluate OpenClaw securely by isolating the AI agent in ephemeral containers. This approach ties into a broader trend among tech companies that are embracing sandbox environments to safely test AI tools, mitigating the risks posed to production systems.
OpenClaw sees 21K exposed deployments; Cloudflare offers safe testing option
