# OpenAI 发布 GPT-5.4-Cyber，一款专为防御性网络安全打造的模型

- 来源：The Decoder：AI News（RSS）
- 作者：Maximilian Schreiner
- 发布时间：2026-04-15 17:50
- AIHOT 链接：https://aihot.virxact.com/items/cmnzwaln5010dslwzelmxmjf3
- 原文链接：https://the-decoder.com/openai-releases-gpt-5-4-cyber-a-model-built-specifically-for-defensive-cybersecurity

## AI 摘要

OpenAI 正式发布 GPT-5.4-Cyber，这是一款专门为防御性网络安全场景训练和优化的专用模型。与通用大模型不同，该模型专注于网络安全防御任务，目前访问权限受到严格限制，仅向经过身份验证的安全领域专家开放，普通用户暂无法使用。

## 正文

OpenAI releases GPT-5.4-Cyber, a model built specifically for defensive cybersecurity

Key Points

OpenAI has released GPT-5.4-Cyber, a model variant fine-tuned for defensive cybersecurity that enables tasks like binary reverse engineering. Access is initially limited to a few hundred verified security professionals.

The model competes directly with Anthropic's Claude Mythos, which was unveiled a week earlier and specializes in finding vulnerabilities in operating systems and browsers.

OpenAI also points to its Codex Security tool, which has reportedly helped patch more than 3,000 critical vulnerabilities, along with a $10 million grant program for cybersecurity projects.

OpenAI has released GPT-5.4-Cyber, a model fine-tuned specifically for defensive cybersecurity work. For now, access is limited to verified security professionals.

OpenAI is expanding its "Trusted Access for Cyber" (TAC) program with a new model built specifically for cybersecurity: GPT-5.4-Cyber. According to OpenAI, it's a variant of GPT-5.4 that's less restrictive when it comes to defensive security work, enabling tasks like binary reverse engineering - the analysis of compiled software without access to source code.

A few hundred users will get access first, with the program expanding to thousands of verified individuals and hundreds of teams over the coming weeks. Because the model is more permissive, it comes with tighter controls: access through third-party platforms or under zero-data-retention agreements may be restricted.

Just one week earlier, competitor Anthropic unveiled Claude Mythos, an AI model specialized in finding and exploiting vulnerabilities in operating systems and browsers. Like GPT-5.4-Cyber, Mythos is only available to a restricted group of users.

Codex Security fits into OpenAI's broader cybersecurity push

Beyond the new model, OpenAI points to its existing security tool Codex Security, which automatically identifies vulnerabilities in codebases and suggests fixes. Since launch, the system has helped patch more than 3,000 critical vulnerabilities. OpenAI is also putting $10 million toward its Cybersecurity Grant Program and has reached over 1,000 open-source projects with free security scanning.

Mythos has already caused concern among financial firms and U.S. government agencies, according to Bloomberg. During a meeting with Wall Street executives, Treasury Secretary Scott Bessent and Fed Chair Jerome Powell warned attendees to take the model seriously. The Treasury Department's technology team is reportedly trying to get access to Mythos itself to check its own systems for vulnerabilities.

AI News Without the Hype – Curated by Humans