new high-severity Firefox bugs OpenAI says GPT-6 Cyber found
汇总自3个来源
★ 今日焦点
OpenAI的GPT-6 Cyber找出81个Firefox漏洞和948个Linux利用
OpenAI推出GPT-6 Cyber,供获批的安全团队寻找并验证软件漏洞。它在ExploitGym上得到48.5%,高于GPT-5.6 Cyber的32.1%,相对提升51%。使用权限先开放给获批的Daybreak OpenAI's programme that gives approved security teams stronger AI tools to find and fix software bugs. Red客户。
风险评级:OpenAI在自有Preparedness Framework OpenAI's own scale for how risky a model's skills are in areas such as cyber attacks.里把网络安全能力评为Critical,高于上一代Cyber的High。它同时加强滥用监测,并为关键基础设施防御方留出通道。
谁能用:GPT-6 Cyber先在API Application programming interface: a door that lets one piece of software ask another for data or actions.和Codex OpenAI's AI coding helper, which can write, fix and test software for you.向精选Daybreak Red开放,再逐步扩大。相关防御工具的Daybreak Blue也在扩面。
政府工具:GovTech A government unit that builds digital tools for public services. Malaysia将推出使用AWS的Kiro An AWS coding tool that helps developers write and change software with AI help.编码工具的GovHub。2026年先试点500名开发者,目标2028年达到4,000人。AWS还将投资,让最多44.5万名公务员用上工作助手Amazon Quick Amazon's workplace AI assistant that helps with research, summaries and data questions.,处理采购数据分析和政策研究等。
微软柔佛:Southeast Asia 3属于微软约89亿令吉(22亿美元)四年马来西亚计划。备用发电机将使用可再生生物燃料。微软称到2025年10月,AIForMYFuture已培训逾73.4万人,目标2027年达到80万。数字部长Gobind Singh Deo也敦促谷歌610亿令吉计划要催生马来西亚技术,而不只是算力。
AWS launches Semarak AI Malaysia, putting Amazon Quick tool to 445,000 government officers · TechNode GlobalRead ↗
Microsoft to open Southeast Asia 3 cloud region in Johor Bahru, per $2.2B Malaysia plan · TechNode GlobalRead ↗
Gobind: Malaysia must turn Google's US$15bil investment into homegrown AI innovation · Digital News AsiaRead ↗
这篇报道有用吗?轻点即可投票。长按、右键或按 Shift+Enter 可先添加理由或备注。
企业如何用AI
2,600
main tables Nitori had to make AI-readable
企业如何用AI
日本宜得利发现:数据没有人读得懂的名字,AI分析就没用
家具零售商宜得利把销售数据迁入Google BigQuery Google's cloud database made for analysing very large sets of business data.后,主要表约有2,600张。懂数据的员工用得顺,其他人却迷路。解决办法不是更聪明的聊天机器人,而是用白话写清每一列的含义。
Nitori: how naming rules blocked conversational BigQuery analysis, and how labels fixed it · ITmedia EnterpriseRead ↗
这篇报道有用吗?轻点即可投票。长按、右键或按 Shift+Enter 可先添加理由或备注。
AI安全
0 clicks
needed for a website to plant the fake instruction file
AI安全
下载文件夹里的假AGENTS.md可在零点击下劫持编程代理
安全公司Forever Security称在真实企业电脑上发现新型攻击。网站悄悄把假AGENTS.md A text file in a folder that tells AI coding agents how to behave there. Agents read it automatically.放进Downloads。信任该文件夹的编程代理会改写自己的设置,并为恶意软件开门。
主要收获
原理:Codex OpenAI's AI coding helper, which can write, fix and test software for you.和Anthropic Claude Code会读取AGENTS.md或CLAUDE.md作为指令。浏览器也能不经点击把文件存进Downloads。两者叠加时,植入的Markdown会像程序一样被执行。
假文件做了什么:一份真实样本让Codex修改~/.codex/config.toml并安装名为function-color的恶意npm package A shareable piece of JavaScript code that developers install into a project; fake ones can be malware.包。一份CLAUDE.md测试约有一半概率改写.claude/settings.json,把流量导向攻击者服务器。
Fake AGENTS.md in the Downloads folder can hijack coding agents without a click · ITmedia EnterpriseRead ↗
这篇报道有用吗?轻点即可投票。长按、右键或按 Shift+Enter 可先添加理由或备注。
AI产品
50%
faster image generation vs Images 2.5, OpenAI says
汇总自2个来源
AI产品
ChatGPT Images 2.6提速50%,还能做GIF
OpenAI于11月4日向所有ChatGPT和Codex OpenAI's AI coding helper, which can write, fix and test software for you.方案(含免费)发布ChatGPT Images 2.6。它称生成比Images 2.5快约50%,编辑时更能保住角色细节,GIF动画正在分批上线。
开发者:API Application programming interface: a door that lets one piece of software ask another for data or actions.提供更快的Flare和更高质量的Sunburst,还有SVG导出,以及控制模型绘制前思考多少的complexity参数。