Adversa AI Releases Open-Source Security Tool

- Adversa AI, the company behind SecureClaw, is a Tel Aviv-based startup founded in 2021 that specializes in AI security, offering services like "ethical hacking" for AI systems and continuous AI Red Teaming. - SecureClaw provides protection against a range of threats to AI agents, including prompt manipulation, credential theft, malicious skills, and agent takeover attempts. These autonomous AI agents introduce unique security vulnerabilities like prompt injection, data leakage, and model poisoning that traditional security tools are not equipped to handle. - The tool offers a layered defense by combining 55 automated audit and hardening checks with a behavioral security skill for runtime protection. - SecureClaw's alignment with the OWASP Top 10 for Agentic Applications is significant as this set of standards addresses the highest impact security risks in autonomous, tool-using, multi-agent systems, which are not adequately covered by traditional web application security standards. - The release of SecureClaw is timely, as the adoption of agentic AI is rapidly increasing, expanding the potential attack surface for enterprises. Experts warn that defensive controls need to evolve to manage risks associated with AI agents accessing files, credentials, and external tools. - Before SecureClaw, Adversa AI also released the MCP Security TOP 25 Framework, a resource that catalogs vulnerabilities and defenses for the Model Context Protocol, which is foundational to how AI agents interoperate. - The open-source nature of SecureClaw allows for community collaboration and transparency, a common theme among several new AI security tools designed to address emerging threats in the AI development lifecycle. - Adversa AI is led by founder and CEO Alex Polyakov and is recognized by industry authorities like Gartner for its work in AI security.