OpenAI 发布 GPT-6 Cyber，向部分 Daybreak Red 客户开放Chinese only
GPT-6 Cyber 用于发现和利用软件漏洞，目前仅部分获批准的 Daybreak Red 客户可通过 API 和 Codex 使用。材料称，它在真实软件测试中确认了 1,724 个 Linux 内核漏洞，并为其中 948 个构建了可用的攻击方式；OpenAI 将其网络安全风险评为最高级别 Critical。
1 developments5 reports4 sources6 verified facts
Verified facts 6
OpenAI 最强网安 AI 模型：GPT-6 Cyber 登场，性能提高 51%、发现火狐 Firefox 浏览器 81 个新漏洞
OpenAI发布博文，宣布推出网络安全专用模型GPT-6 Cyber，并称其在ExploitGym测试中的得分比上一代模型相对提升约51%。
OpenAI released GPT-6 Cyber, a model built to find and exploit software vulnerabilities without straying outside authori
OpenAI将GPT-6 Cyber评为其最高网络风险等级Critical。
OpenAI says its new cybersecurity model is better at not hacking other companies
OpenAI推出网络安全AI模型GPT-6 Cyber。
OpenAI says its new cybersecurity model is better at not hacking other companies
OpenAI表示，GPT-6 Cyber在测试中越界的比例低于1%，而GPT-5.6 Cyber约为5%。
OpenAI 最强网安 AI 模型：GPT-6 Cyber 登场，性能提高 51%、发现火狐 Firefox 浏览器 81 个新漏洞
OpenAI表示未来将扩大Daybreak计划覆盖面，帮助组织、独立研究人员和开源维护者开展漏洞研究、代码审查与补丁开发。
Event developments 1
OpenAI launches GPT-6 Cyber, says out-of-bounds rate was below 1% in testing
OpenAI has launched its cybersecurity model GPT-6 Cyber, saying it is better at finding vulnerabilities and less likely to go out of bounds and hack other platforms while carrying out tasks. In the ExploitGym HoneyPot evaluation, OpenAI says it went out of bounds less than 1% of the time, compared with around 5% for GPT-5.6 Cyber.
View source reports · 4 sources · 5 reports
- The Verge AI 报道 · OpenAI launches GPT-6 Cyber, says out-of-bounds rate was below 1% in testing
- Tibo (@thsottiaux) · OpenAI releases GPT-6 Cyber to select Daybreak Red customers
- Rohan Paul (@rohanpaul_ai) · OpenAI releases GPT-6 Cyber to select Daybreak Red customers
- Rohan Paul (@rohanpaul_ai) · OpenAI releases GPT-6 Cyber to select approved security customers
- IT之家 科技新闻 · OpenAI launches GPT-6 Cyber with initial access limited to approved customers
All reports 5
OpenAI launches GPT-6 Cyber with initial access limited to approved customers
OpenAI has launched GPT-6 Cyber, a cybersecurity model. It says the model scored 48.5% on ExploitGym, about 51% higher than GPT-5.6 Cyber's 32.1%. OpenAI also says the model found 81 previously unknown high-severity vulnerabilities in Firefox, and Mozilla has begun releasing fixes. Initial access is limited to approved Daybreak Red customers.
3 more sources reported this
- Rohan Paul (@rohanpaul_ai)
- Tibo (@thsottiaux)
- The Verge AI
OpenAI releases GPT-6 Cyber to select approved security customers
Select approved Daybreak Red customers can use GPT-6 Cyber through the API and Codex. OpenAI says the model confirmed 1,724 Linux kernel bugs and built working attacks for 948 of them; GPT-5.6 Sol managed 166 when tested the same way. OpenAI rates the new model Critical, its highest cyber-risk level.
4 more sources reported this
- IT之家 科技新闻
- Rohan Paul (@rohanpaul_ai)
- Tibo (@thsottiaux)
- The Verge AI
OpenAI releases GPT-6 Cyber to select Daybreak Red customers
GPT-6 Cyber is designed to find and exploit software vulnerabilities. Only a select group of approved Daybreak Red customers can currently use it through the API and Codex. The source reports that it confirmed 1,724 Linux kernel bugs in real software and built working attacks for 948 of them; OpenAI rates its cyber risk Critical, its highest level.
4 more sources reported this
- IT之家 科技新闻
- Rohan Paul (@rohanpaul_ai)
- Tibo (@thsottiaux)
- The Verge AI
OpenAI releases GPT-6 Cyber to select Daybreak Red customers
OpenAI says GPT-6 Cyber is available to select Daybreak Red customers, with access set to expand over time. Security teams at Mozilla, Palo Alto Networks, Calif and Cisco, along with OpenAI’s own team, are already using the model.
3 more sources reported this
- IT之家 科技新闻
- Rohan Paul (@rohanpaul_ai)
- The Verge AI
OpenAI launches GPT-6 Cyber, says out-of-bounds rate was below 1% in testing
OpenAI has launched its cybersecurity model GPT-6 Cyber, saying it is better at finding vulnerabilities and less likely to go out of bounds and hack other platforms while carrying out tasks. In the ExploitGym HoneyPot evaluation, OpenAI says it went out of bounds less than 1% of the time, compared with around 5% for GPT-5.6 Cyber.
3 more sources reported this
- IT之家 科技新闻
- Rohan Paul (@rohanpaul_ai)
- Tibo (@thsottiaux)