SAN FRANCISCO: OpenAI has unveiled Aardvark, a new AI assistant designed to help security teams identify and fix vulnerabilities, powered by the latest GPT-5 technology. Released on Thursday, the agent aims to address the growing number of software security issues, with tens of thousands of new vulnerabilities reported each year in both enterprise and open-source code.
Initially developed as an internal tool for OpenAI’s own developers, Aardvark has shown promise in explaining security flaws clearly and guiding teams to actionable fixes. OpenAI’s VP Matt Knight emphasized that the agent’s reasoning and tool-assisted approach could bring meaningful improvements to software security workflows.
Read More: OpenAI unveils GPT-5 with smarter AI and coding boost
Aardvark operates through a multi-stage process: it examines a code repository to understand its purpose, scans for potential vulnerabilities, annotates code for human review, and tests suspected issues in a sandboxed environment. Once verified, it leverages OpenAI’s Codex coding assistant to propose patches, which are then reviewed and implemented by human developers.
The tool is currently available in private beta for select partners invited by OpenAI. Feedback from participants will help refine its detection accuracy, validation workflows, and overall functionality before a wider rollout.
Read More: Anthropic unveils claude sonnet 4.5 AI breakthrough
OpenAI’s introduction of Aardvark highlights the growing role of AI in cybersecurity, offering teams a scalable and intelligent way to manage vulnerabilities while complementing human oversight.