OpenAI to launch Codex products next week, Altman says

OpenAI to launch Codex products next week, Altman says

by on 24/01/2026 with No Comments

OpenAI CEO Sam Altman announced that the company plans to release several Codex-related products in the coming month, with the first launch scheduled for next week. In a post on X, Altman revealed that OpenAI is approaching the "Cybersecurity High level" on its internal preparedness framework, signaling a major milestone in the development of AI-powered coding tools.

The announcement comes as OpenAI prepares to implement product restrictions designed to prevent misuse of its increasingly powerful coding models for criminal activities such as unauthorized system intrusions and financial theft.

Restrictions and Defensive Strategy

Altman addressed the dual-use nature of cybersecurity tools, explaining that OpenAI will initially implement safeguards to block users from employing the technology for malicious purposes. "These restrictions will attempt to block users from employing the technology for criminal activities, such as hacking into banks to steal money," Altman wrote.

The company's long-term strategy involves shifting toward what it calls "defensive acceleration," which prioritizes helping users discover and patch security vulnerabilities rather than restricting capabilities entirely. Altman indicated this approach would be implemented as OpenAI gathers evidence supporting its effectiveness.

"It is very important the world adopts these tools quickly to make software more secure," Altman wrote, noting that "there will be many very capable models in the world soon."

Rising Capabilities and Preparedness

Under OpenAI's Preparedness Framework, reaching "High" cybersecurity capability means a model could potentially develop working zero-day remote exploits against well-defended systems or assist with complex, stealthy intrusion operations targeting enterprise or industrial networks. OpenAI has confirmed it is planning and testing future systems as if each new model could reach that level.

The company's cybersecurity capabilities have improved sharply in recent months. Capabilities assessed through capture-the-flag challenges improved from 27% on GPT-5 in August 2025 to 76% on GPT-5.1-Codex-Max in November 2025, according to OpenAI.

To manage dual-use risks, OpenAI has outlined a defense-in-depth approach combining access controls, hardened infrastructure, egress limitations, and monitoring. The company is also training its models to decline requests that could enable cyber abuse while remaining helpful for legitimate defensive purposes.

Broader Security Ecosystem

OpenAI has been expanding its defensive cybersecurity tools. Aardvark, the company's autonomous security researcher that scans codebases and proposes vulnerability patches, is now in private beta and has already identified novel Common Vulnerabilities and Exposures in open-source projects. The company is also establishing a Frontier Risk Council to advise on balancing useful capabilities with potential misuse concerns.

The upcoming Codex launches follow a period of rapid product development, with OpenAI having released GPT-5.2-Codex in December and beginning to roll out a "Codex-Max" variant to paid users earlier this month.