OpenAI launched the AI agent Codex Security, capable of independently detecting and fixing software vulnerabilities

OpenAI launches a new AI agent for cybersecurity

OpenAI introduced Codex Security—a specialized agent focused on detecting and mitigating threats in the field of information security. According to the company, the tool can process large volumes of data and automatically generate correct patches, thereby freeing developers from routine tasks.

Technical foundation

Bloomberg reports that Codex Security is built on the Aardvark project, which used the GPT‑5 model in closed testing by OpenAI last year. Now this technology has become a commercial product capable of scanning code repositories and improving its algorithms under user feedback.

Market impact

The announcement increases pressure on traditional information security solution providers and confirms the automation trend in protection that Anthropic started. Last month, Anthropic’s Claude Code Security tool caused Crowdstrike and Cloudflare shares to fall by about 8 %, demonstrating market sensitivity to such innovations.

Availability

Codex Security will be available in the coming days as a research preview for corporate customers using ChatGPT Enterprise and Business, as well as educational institutions. OpenAI offers a free one‑month trial so organizations can evaluate the system’s effectiveness without initial investment.