CertiK Unveils AI Skill Scanner to Detect Hidden Security Risks

• CertiK Skill Scanner is designed to assess risks that may arise during actual execution, such as situations requiring fund calls and financial transactions.
• Before letting third-party AI skills into internal or production settings, businesses may use the scanner as part of internal compliance and risk management procedures.

Today, CertiK, a blockchain and AI security firm, officially launched CertiK Skill Scanner, a new security solution intended to safeguard the quickly expanding ecosystem of AI Agents and third-party AI Skills.

The solution is designed for developers, businesses, AI skill markets, and ultimately regular consumers who want more insight into the dangers of using third-party AI technologies. According to CertiK, the platform is meant to assist in addressing rising concerns about risks associated with autonomous execution in AI contexts, disguised harmful behavior, and unauthorized data access.

CertiK Skill Scanner is designed to assess risks that may arise during actual execution, such as situations requiring fund calls and financial transactions, in contrast to broader AI scanning tools. Both Web3 and conventional Web2 ecosystems may use the product.

AI skill marketplaces can automatically analyze skills before they go online by integrating CertiK Skill Scanner directly into publishing pipelines. CertiK security verdicts may also be shown on marketplaces as trust indicators for end consumers assessing third-party skills.

Before letting third-party AI skills into internal or production settings, businesses may use the scanner as part of internal compliance and risk management procedures.

Before publishing, independent developers may self-audit their skills using the scanner. Upcoming product upgrades will increase direct access for regular users, allowing them to check Skills before installing or using them.

Along with “pass,” “warn,” or “fail” judgments and a limited findings list organized by severity, the scanner produces a scored evaluation that ranges from 0 to 100. CertiK claims that the system can detect security threats with up to 90.5% accuracy, which helps lower false positives and increase the accuracy of AI skill risk evaluations.

A few Web3 AI Agent infrastructure settings have already implemented the offering. Additionally, CertiK is developing partnerships with other AI Skill platforms, such as FinChip.ai.

CertiK’s further growth into AI-focused security infrastructure comes after the launch. In an attempt to address new risks associated with autonomous systems and AI-driven execution environments, the business launched its AI Auditor program earlier this year.