The ChatGPT-maker’s product rivals Anthropic PBC’s Claude Code Security, which also spots vulnerabilities and suggests solutions to mend them.
| Photo Credit:
Dado Ruvic
OpenAI is introducing an AI agent that’s meant to help security teams find and patch vulnerabilities in large databases, potentially cutting into demand for legacy cyber firms.
The agent, called Codex Security, works by identifying cybersecurity flaws and proposing solutions before fixing the bugs, OpenAI said in a statement Friday. The tool is designed to “operate at scale” and provide “easy-to-accept patches,” and thus enable developers to focus on higher-level tasks, the company said. The San Francisco-based firm also shared that the agentic tool has been used to scan and identify security vulnerabilities within open-source repositories.
The ChatGPT-maker’s product rivals Anthropic PBC’s Claude Code Security, which also spots vulnerabilities and suggests solutions to mend them. The release of Caude’s new security tool last month sent cybersecurity stocks tumbling. Both Crowdstrike Holdings and Cloudflare Inc fell by 8 per cent at the time.
Codex Security evolved from Aardvark, an agentic security researcher powered by GPT‑5 which OpenAI released last year to beta test on a small cohort of customers. Codex Security will be rolled out in the following days as a research preview to ChatGPT Enterprise, Business and college customers with the first month being free.
The agent is designed to learn from feedback to improve its threat models over time.
More stories like this are available on bloomberg.com
Published on March 7, 2026
Source link
#OpenAI #releases #agent #security #tool #research #preview
