Story perspectives

OpenAI Limits GPT-5.4-Cyber Release Amid Security Concerns

4/15/2026

Summary

Story summary

OpenAI will limit the release of GPT-5.4-Cyber to a select group of partners, following Anthropic's similar decision over cybersecurity concerns.
OpenAI aims to provide defensive tools to organizations protecting critical infrastructure.
The technology identifies security vulnerabilities and can be used offensively as well as defensively.
OpenAI board member Zico Kolter emphasized a cautious approach amid recent AI advancements.
The release strategy aims to give defenders an advantage over potential attackers.

1 of 1

Story summary

OpenAI will limit the release of GPT-5.4-Cyber to a select group of partners, following Anthropic's similar decision over cybersecurity concerns.
OpenAI aims to provide defensive tools to organizations protecting critical infrastructure.
The technology identifies security vulnerabilities and can be used offensively as well as defensively.
OpenAI board member Zico Kolter emphasized a cautious approach amid recent AI advancements.
The release strategy aims to give defenders an advantage over potential attackers.