https://t.me/AI_News_CN
📈主流AI服务状态页通知 | 🆕汇集全网ChatGPT/AI新闻 #AI #ChatGPT
🆓免费AI聊天 https://free.netfly.top
✨BEST AI中转 https://api.oaibest.com 2.8-4.2折 支持OpenAI, Claude, Gemini,Grok, Deepseek, Midjourney, 文件上传分析
Buy ads: https://telega.io/c/AI_News_CN
📈主流AI服务状态页通知 | 🆕汇集全网ChatGPT/AI新闻 #AI #ChatGPT
🆓免费AI聊天 https://free.netfly.top
✨BEST AI中转 https://api.oaibest.com 2.8-4.2折 支持OpenAI, Claude, Gemini,Grok, Deepseek, Midjourney, 文件上传分析
Buy ads: https://telega.io/c/AI_News_CN
Intermittent errors connecting to Claude.ai Connectors
Jul 17, 23:29 UTC
Identified - We have identified the cause of the issue affecting connecting to Claude.ai Connectors. We are deploying a fix. We will provide a further update shortly.
via Anthropic Status - Incident History
Jul 17, 23:29 UTC
Identified - We have identified the cause of the issue affecting connecting to Claude.ai Connectors. We are deploying a fix. We will provide a further update shortly.
via Anthropic Status - Incident History
OpenAI今日宣布推出ChatGPT的全新“代理模式”,该模式将ChatGPT从一个信息提供者转变为一个能够自主执行复杂任务的智能代理。用户现在可以指示ChatGPT在虚拟计算机上完成从头到尾的工作,例如:
* 日程管理与信息整合: “查看我的日历,并根据近期新闻为我简报即将举行的客户会议。”
* 生活服务自动化: “计划并购买制作四人份日式早餐所需的食材。”
* 专业分析与报告生成: “分析三家竞争对手,并创建一个演示文稿。”
新推出的代理能够智能地浏览网站、过滤信息、在必要时提示用户安全登录、运行代码、进行分析,并交付可编辑的演示文稿和电子表格。这一能力整合了Operator(网页交互)、深度研究(信息综合)以及ChatGPT自身的智能和对话流畅性。
关键能力与优势:
* 统一的代理系统: 结合了网页交互、深度研究和对话智能,实现端到端的任务处理。
* 多工具支持: 拥有视觉浏览器、文本浏览器、终端和直接API访问,并可连接Gmail、GitHub等应用。
* 用户全程可控: 用户可以随时中断、接管浏览器或停止任务,ChatGPT在执行重要操作前会请求许可。
* 广泛的实际应用: 无论是自动化工作中的重复性任务(如将图表转换为演示文稿、更新电子表格),还是个人生活中的行程规划、活动组织,都能提供强大支持。
性能表现亮眼:
在多项基准测试中,ChatGPT代理展现出卓越的性能:
* 在“人类最后的考试”(Humanity’s Last Exam)中,pass@1得分达到41.6,通过并行策略可提升至44.4。
* 在最难的数学基准测试FrontierMath中,准确率达到27.4%,远超先前模型。
* 在模拟复杂现实世界知识工作任务的内部基准测试中,其产出与人类相当或更优,尤其在数据科学任务(DSBench)和电子表格编辑(Spre
(HackerNews)
via 茶馆 - Telegram Channel
Elevated network errors while using ChatGPT
Status: Investigating
We are investigating the issue for the listed services.
Affected components
● Mobile - iOS (Degraded performance)
● Mobile - Android (Degraded performance)
● Web (Degraded performance)
● Desktop - macOS (Degraded performance)
● Desktop - Windows (Degraded performance)
via OpenAI status
Status: Investigating
We are investigating the issue for the listed services.
Affected components
● Mobile - iOS (Degraded performance)
● Mobile - Android (Degraded performance)
● Web (Degraded performance)
● Desktop - macOS (Degraded performance)
● Desktop - Windows (Degraded performance)
via OpenAI status
Statement from the OpenAI Board of Directors on the Nonprofit Commission Report
董事会感谢独立的OpenAI非营利委员会成员们的辛勤工作和积极参与。该委员会由OpenAI于四月召集[https://openai.com/index/nonprofit-commission-advisors/],旨在支持“一项参与流程,从广泛的利益相关者中收集经验和反馈,探讨OpenAI的慈善事业如何解决长期的系统性问题”,并提供相关建议。随附的报告反映了委员会的独立调查结果。
我们感谢为委员会报告提供专业知识的各方,也衷心感谢社区领袖和一线从业者在整个过程中慷慨分享他们的观点。他们的贡献为OpenAI非营利组织如何最好地实现其使命——确保通用人工智能惠及全人类——提供了宝贵的参考。
在推进这一使命的过程中,我们将继续致力于倾听、学习,并与变革前线的伙伴携手共建,作为董事会对一个资源充足、使命驱动且响应其服务社区需求的非营利组织的承诺的一部分。
----------------------
The Board of Directors thanks the members of the independent OpenAI Nonprofit Commission for their extensive work and engagement. The Commission was convened by OpenAI in April to support “an engagement process to gather learnings and feedback from a wide range of stakeholders on how OpenAI’s philanthropy can address long-term systemic issues” and provide recommendations. The attached report reflects the Commission’s independent findings.
We appreciate the expertise that informed the Commission’s report, and are grateful to the community leaders and frontline practitioners who generously shared their perspectives throughout the process. Their contributions provide valuable input into how OpenAI’s nonprofit can best fulfill its mission of ensuring Artificial General Intelligence benefits all of humanity.
As we carry this mission forward, we remain committed to listening, learning, and building in partnership with those on the front lines of change, as part of the Board’s commitment to a nonprofit that is well-resourced, mission-led, and responsive to the needs of the communities it aims to serve.
via OpenAI News
董事会感谢独立的OpenAI非营利委员会成员们的辛勤工作和积极参与。该委员会由OpenAI于四月召集[https://openai.com/index/nonprofit-commission-advisors/],旨在支持“一项参与流程,从广泛的利益相关者中收集经验和反馈,探讨OpenAI的慈善事业如何解决长期的系统性问题”,并提供相关建议。随附的报告反映了委员会的独立调查结果。
我们感谢为委员会报告提供专业知识的各方,也衷心感谢社区领袖和一线从业者在整个过程中慷慨分享他们的观点。他们的贡献为OpenAI非营利组织如何最好地实现其使命——确保通用人工智能惠及全人类——提供了宝贵的参考。
在推进这一使命的过程中,我们将继续致力于倾听、学习,并与变革前线的伙伴携手共建,作为董事会对一个资源充足、使命驱动且响应其服务社区需求的非营利组织的承诺的一部分。
----------------------
The Board of Directors thanks the members of the independent OpenAI Nonprofit Commission for their extensive work and engagement. The Commission was convened by OpenAI in April to support “an engagement process to gather learnings and feedback from a wide range of stakeholders on how OpenAI’s philanthropy can address long-term systemic issues” and provide recommendations. The attached report reflects the Commission’s independent findings.
We appreciate the expertise that informed the Commission’s report, and are grateful to the community leaders and frontline practitioners who generously shared their perspectives throughout the process. Their contributions provide valuable input into how OpenAI’s nonprofit can best fulfill its mission of ensuring Artificial General Intelligence benefits all of humanity.
As we carry this mission forward, we remain committed to listening, learning, and building in partnership with those on the front lines of change, as part of the Board’s commitment to a nonprofit that is well-resourced, mission-led, and responsive to the needs of the communities it aims to serve.
via OpenAI News
OpenAI 今日宣布推出 ChatGPT 智能体功能,该功能整合了此前 Operator 和深度研究工具的能力,可在虚拟计算机环境中自主完成复杂任务。用户可要求其处理日程安排、竞品分析、制作演示文稿等工作流程。
新功能配备可视化浏览器、文本浏览器、终端和 API 访问等工具套件,支持网站交互、数据分析和文件处理。该功能已在多项学术评估中达行业领先水平。Pro 用户每月可使用 400 次,Plus 和 Team 用户为 40 次。该功能今日起向付费用户分批开放,企业和教育用户将在未来几周内获得访问权限。
OpenAI | OpenAI(中文稿)| Livestream
🍀频道 🍵茶馆 📮投稿
Video is too big
via 科技圈🎗在花频道📮 - Telegram Channel
Elevated Error Rates on Anthropic Models
Jul 17, 21:13 UTC
Investigating - We are investigating this issue.
via Cursor Status - Incident History
Jul 17, 21:13 UTC
Investigating - We are investigating this issue.
via Cursor Status - Incident History
Elevated errors for requests to Claude Sonnet and Opus models
Jul 17, 21:13 UTC
Update - We are continuing to investigate this issue.
Jul 17, 21:13 UTC
Update - We are continuing to investigate this issue.
Jul 17, 21:13 UTC
Investigating - We are currently investigating elevated errors on requests to Claude Sonnet (all models) on the API, Claude.ai, and the Anthropic Console.
via Anthropic Status - Incident History
Jul 17, 21:13 UTC
Update - We are continuing to investigate this issue.
Jul 17, 21:13 UTC
Update - We are continuing to investigate this issue.
Jul 17, 21:13 UTC
Investigating - We are currently investigating elevated errors on requests to Claude Sonnet (all models) on the API, Claude.ai, and the Anthropic Console.
via Anthropic Status - Incident History
🤖 ChatGPT推出“Agent”功能,可自主登录账户执行任务
OpenAI近日发布了其AI聊天机器人ChatGPT的一项新功能——“Agent”。该功能允许ChatGPT在用户的授权下,自主登录第三方服务并执行一系列任务,例如检查日历、规划膳食并订购杂货等。
核心亮点:
* 自主任务执行: Agent能够代表用户完成从头到尾的任务,甚至可以登录用户的账户(如GrubHub、Google Drive、GitHub、SharePoint等)来完成这些任务。
* 付费订阅: 此功能仅对每月订阅费为20美元的ChatGPT Plus用户开放。
* 文件创建能力: Agent还可以直接在Microsoft Excel和PowerPoint中创建电子表格和演示文稿,并将其发送给用户下载。
* 安全性考量: OpenAI表示,Agent在执行重要操作前会请求用户许可,并且用户可以随时中断任务。然而,建议用户谨慎提供敏感账户凭证。
* 技术整合: 该功能整合了OpenAI的Operator、Deep Research和ChatGPT的对话能力,旨在提升用户搜索和执行任务的效率。
此举标志着AI在自动化和用户服务方面迈出了重要一步,同时也可能加剧OpenAI与微软等科技巨头在AI服务领域的竞争。
(PCMag.com)
via 茶馆 - Telegram Channel
OpenAI近日发布了其AI聊天机器人ChatGPT的一项新功能——“Agent”。该功能允许ChatGPT在用户的授权下,自主登录第三方服务并执行一系列任务,例如检查日历、规划膳食并订购杂货等。
核心亮点:
* 自主任务执行: Agent能够代表用户完成从头到尾的任务,甚至可以登录用户的账户(如GrubHub、Google Drive、GitHub、SharePoint等)来完成这些任务。
* 付费订阅: 此功能仅对每月订阅费为20美元的ChatGPT Plus用户开放。
* 文件创建能力: Agent还可以直接在Microsoft Excel和PowerPoint中创建电子表格和演示文稿,并将其发送给用户下载。
* 安全性考量: OpenAI表示,Agent在执行重要操作前会请求用户许可,并且用户可以随时中断任务。然而,建议用户谨慎提供敏感账户凭证。
* 技术整合: 该功能整合了OpenAI的Operator、Deep Research和ChatGPT的对话能力,旨在提升用户搜索和执行任务的效率。
此举标志着AI在自动化和用户服务方面迈出了重要一步,同时也可能加剧OpenAI与微软等科技巨头在AI服务领域的竞争。
(PCMag.com)
via 茶馆 - Telegram Channel
无需WSL!直接在Windows CMD 中使用Claude Code
https://mp.weixin.qq.com/s/LLkeDY5pvfB9ncz6hg-6Nw
https://mp.weixin.qq.com/s/LLkeDY5pvfB9ncz6hg-6Nw
OpenAI将向Pro Plus和Team用户推出ChatGPT Agent
新版ChatGPT智能体,这是一个统一的人工智能代理,能够利用其自有的虚拟计算机和各种工具来执行复杂的任务,该智能体旨在处理复杂任务,能够在思考、执行操作和使用各种工具之间无缝切换,它可以使用终端、在网络上进行点击操作,并生成电子表格和幻灯片等文档
该功能结合了Operator和deep research,新的智能体解决了先前工具的局限性,例如操作员处理长篇文章的困难,以及深度研究在交互式网页或登录需要身份验证的来源方面的限制,该智能体被设计为协作性的,允许用户提供初始指令、澄清问题,并在任务中途提供进一步的指导,它可以提出澄清性问题,可以被中断,并可能在重要步骤前请求确认,例如在发送电子邮件之前,用户还可以直接接管浏览器以纠正错误或输入敏感信息
该智能体在多个基准上进行了评估:
via LoopDNS资讯播报 - Telegram Channel
新版ChatGPT智能体,这是一个统一的人工智能代理,能够利用其自有的虚拟计算机和各种工具来执行复杂的任务,该智能体旨在处理复杂任务,能够在思考、执行操作和使用各种工具之间无缝切换,它可以使用终端、在网络上进行点击操作,并生成电子表格和幻灯片等文档
该功能结合了Operator和deep research,新的智能体解决了先前工具的局限性,例如操作员处理长篇文章的困难,以及深度研究在交互式网页或登录需要身份验证的来源方面的限制,该智能体被设计为协作性的,允许用户提供初始指令、澄清问题,并在任务中途提供进一步的指导,它可以提出澄清性问题,可以被中断,并可能在重要步骤前请求确认,例如在发送电子邮件之前,用户还可以直接接管浏览器以纠正错误或输入敏感信息
该智能体在多个基准上进行了评估:
via LoopDNS资讯播报 - Telegram Channel
#Update #ChatGPT
ChatGPT Agent Mode 现已正式发布:
- 可以自动化访问网页、点击网页元素、运行代码、创建 Artifact、执行 Deep Research
- 可以自主生成图片并创建 PowerPoint 文件
- 提供可视化界面,直观看到 Agent 的操作
- 宣称多项性能超越了连接了 Tools 的 o3 模型
接下来几天内所有订阅用户可用,Pro 用户每月 400 次,其他订阅用户每月 40 次。
此外,现在也支持通过额外付费来获得更多使用额度。
via AI Copilot - Telegram Channel
ChatGPT Agent Mode 现已正式发布:
- 可以自动化访问网页、点击网页元素、运行代码、创建 Artifact、执行 Deep Research
- 可以自主生成图片并创建 PowerPoint 文件
- 提供可视化界面,直观看到 Agent 的操作
- 宣称多项性能超越了连接了 Tools 的 o3 模型
接下来几天内所有订阅用户可用,Pro 用户每月 400 次,其他订阅用户每月 40 次。
此外,现在也支持通过额外付费来获得更多使用额度。
via AI Copilot - Telegram Channel
🎥 AI生成视频技术突破,内容创作迎来新纪元
近日,一项AI生成视频技术取得重大突破,能够根据文本描述生成高质量的视频内容。该技术展示了其在内容创作领域的巨大潜力,预示着未来媒体和娱乐行业将发生深刻变革。
(@OpenAI)
via 茶馆 - Telegram Channel
近日,一项AI生成视频技术取得重大突破,能够根据文本描述生成高质量的视频内容。该技术展示了其在内容创作领域的巨大潜力,预示着未来媒体和娱乐行业将发生深刻变革。
(@OpenAI)
via 茶馆 - Telegram Channel
ChatGPT agent System Card
ChatGPT 代理是一种新型的智能代理模型,属于与 OpenAI o3 同一系列,结合了 Deep Research 和 Operator 的优势。它整合了:
● Deep Research 进行多步骤研究并生成高质量报告的能力
● Operator 通过远程可视化浏览器环境执行任务的能力
● 具有限制网络访问权限的终端工具,用于执行代码、进行数据分析以及生成幻灯片或电子表格
● 通过一方连接器访问外部数据源和应用(例如 Google Drive)
从一开始,我们就将安全性作为系统的内在部分,基于 Operator 研究预览中的强大控制措施,增加了额外的安全防护,以应对更广泛用户覆盖和终端访问等新风险。有关详细信息,请参见下文的“产品特定风险缓解”部分。
我们决定根据我们的准备框架,将此次发布视为生物和化学领域的高能力模型[https://openai.com/index/updating-our-preparedness-framework/],并启动相应的安全防护措施。虽然我们没有确凿证据表明该模型能够显著帮助初学者制造严重的生物危害——这是我们为高能力定义的门槛[https://cdn.openai.com/pdf/18a02b5d-6b67-4cec-ab64-68cdfbddebcd/preparedness-framework-v2.pdf]——但我们选择采取预防性措施。
----------------------
ChatGPT agent is a new agentic model in the same family as OpenAI o3, that combines the strengths of deep research and Operator. It brings together:
● Deep Research’s ability to conduct multi-step research and generate high-quality reports
● Operator’s capacity to execute tasks through a remote visual browser environment
● Terminal tool with limited network access for executing code, performing data analysis, and generating slides or spreadsheets
● Access to external data sources and applications (e.g., Google Drive) via first-party Connectors.
From the outset, we've prioritized safety as an inherent part of the system, expanding on robust controls from Operator’s research preview and adding additional safeguards to address new risks like broader user reach and terminal access. See the section on Product-Specific Risk Mitigations below for more detail.
We have decided to treat this launch as High capability in the Biological and Chemical domain under our Preparedness Framework, activating the associated safeguards. While we do not have definitive evidence that this model could meaningfully help a novice to create severe biological harm–our defined threshold for High capability—we have chosen to take a precautionary approach.
via OpenAI News
ChatGPT 代理是一种新型的智能代理模型,属于与 OpenAI o3 同一系列,结合了 Deep Research 和 Operator 的优势。它整合了:
● Deep Research 进行多步骤研究并生成高质量报告的能力
● Operator 通过远程可视化浏览器环境执行任务的能力
● 具有限制网络访问权限的终端工具,用于执行代码、进行数据分析以及生成幻灯片或电子表格
● 通过一方连接器访问外部数据源和应用(例如 Google Drive)
从一开始,我们就将安全性作为系统的内在部分,基于 Operator 研究预览中的强大控制措施,增加了额外的安全防护,以应对更广泛用户覆盖和终端访问等新风险。有关详细信息,请参见下文的“产品特定风险缓解”部分。
我们决定根据我们的准备框架,将此次发布视为生物和化学领域的高能力模型[https://openai.com/index/updating-our-preparedness-framework/],并启动相应的安全防护措施。虽然我们没有确凿证据表明该模型能够显著帮助初学者制造严重的生物危害——这是我们为高能力定义的门槛[https://cdn.openai.com/pdf/18a02b5d-6b67-4cec-ab64-68cdfbddebcd/preparedness-framework-v2.pdf]——但我们选择采取预防性措施。
----------------------
ChatGPT agent is a new agentic model in the same family as OpenAI o3, that combines the strengths of deep research and Operator. It brings together:
● Deep Research’s ability to conduct multi-step research and generate high-quality reports
● Operator’s capacity to execute tasks through a remote visual browser environment
● Terminal tool with limited network access for executing code, performing data analysis, and generating slides or spreadsheets
● Access to external data sources and applications (e.g., Google Drive) via first-party Connectors.
From the outset, we've prioritized safety as an inherent part of the system, expanding on robust controls from Operator’s research preview and adding additional safeguards to address new risks like broader user reach and terminal access. See the section on Product-Specific Risk Mitigations below for more detail.
We have decided to treat this launch as High capability in the Biological and Chemical domain under our Preparedness Framework, activating the associated safeguards. While we do not have definitive evidence that this model could meaningfully help a novice to create severe biological harm–our defined threshold for High capability—we have chosen to take a precautionary approach.
via OpenAI News
Agent bio bug bounty call
邀请函
作为我们持续加强生物领域先进人工智能能力安全保障工作的组成部分,我们的生物漏洞赏金计划现已开放申请。我们已部署了ChatGPT代理模型,并正积极努力进一步强化ChatGPT代理及其他模型的安全防护。我们诚邀具备AI红队测试、安全或化学与生物风险经验的研究人员,尝试寻找一种通用的绕过机制,以破解我们设定的十级生物/化学安全挑战。
项目概述
● 适用模型:仅限ChatGPT代理模型。
● 挑战内容:找到一个通用的绕过提示,能够在全新对话中成功回答所有十个生物/化学安全问题。
● 奖励: • 首个成功通过所有十个问题的真正通用绕过者将获得25,000美元奖金。 • 首个使用多个绕过提示回答所有十个问题的团队将获得10,000美元奖金。 • 对部分成功者,我们可酌情授予较小奖励。
● 时间安排:申请自2025年7月17日开放,采用滚动录取。测试将于2025年7月29日开始。
● 访问权限:申请及参与均需邀请。我们将向经过审核的可信生物红队成员发出邀请,并审查新申请。成功入选者将被引导加入生物漏洞赏金平台。
● 保密协议:所有提示、回复、发现及沟通内容均受保密协议约束。
如何申请
请于2025年7月29日前通过此链接提交简短申请(新窗口打开)[https://openai.smapply.org/prog/bio_bug_bounty],内容包括姓名、所属机构、简要经历及150字的计划说明。被接受的申请者及合作伙伴需拥有现有ChatGPT账户,并签署保密协议。
立即申请,助力我们打造更安全的前沿人工智能。
申请OpenAI生物漏洞赏金计划
申请链接:[https://openai.smapply.org/prog/bio_bug_bounty/]
----------------------
Invitation
As part of our ongoing efforts to strengthen our safeguards for advanced AI capabilities in biology, our bio bug bounty is now open for applications. We’ve deployed the ChatGPT agent model and are actively working to further strengthen safety protections for ChatGPT agent and other models. We’re inviting researchers with experience in AI red teaming, security, or chemical and biological risk to try to find a universal jailbreak that can defeat our ten-level bio/chem challenge.
Program overview
● Model in scope: ChatGPT agent only.
● Challenge: Identify one universal jailbreaking prompt to successfully answer all ten bio/chem safety questions from a clean chat.
● Rewards: • $25,000 to the first true universal jailbreak to clear all ten questions.
• $10,000 to the first team that answers all ten questions with multiple jailbreak prompts.
• Smaller awards may be granted for partial wins at our discretion.
● Timeline: Applications open July 17, 2025 with rolling acceptances. Testing begins July 29, 2025.
● Access: Application and invite-only. We will extend invitations to a vetted list of trusted bio red-teamers and review new applications. Once selected, successful applicants will be onboarded to the bio bug bounty platform.
● Disclosure: All prompts, completions, findings, and communications are covered by NDA.
How to apply
Submit a short application here (name, affiliation, brief track record, and a 150-word plan) by July 29, 2025. Accepted applicants and collaborators must have existing ChatGPT accounts to apply, and will sign a NDA.
Apply now and help us make frontier AI safer.
Apply to the OpenAI bio bug bounty
Apply
via OpenAI News
邀请函
作为我们持续加强生物领域先进人工智能能力安全保障工作的组成部分,我们的生物漏洞赏金计划现已开放申请。我们已部署了ChatGPT代理模型,并正积极努力进一步强化ChatGPT代理及其他模型的安全防护。我们诚邀具备AI红队测试、安全或化学与生物风险经验的研究人员,尝试寻找一种通用的绕过机制,以破解我们设定的十级生物/化学安全挑战。
项目概述
● 适用模型:仅限ChatGPT代理模型。
● 挑战内容:找到一个通用的绕过提示,能够在全新对话中成功回答所有十个生物/化学安全问题。
● 奖励: • 首个成功通过所有十个问题的真正通用绕过者将获得25,000美元奖金。 • 首个使用多个绕过提示回答所有十个问题的团队将获得10,000美元奖金。 • 对部分成功者,我们可酌情授予较小奖励。
● 时间安排:申请自2025年7月17日开放,采用滚动录取。测试将于2025年7月29日开始。
● 访问权限:申请及参与均需邀请。我们将向经过审核的可信生物红队成员发出邀请,并审查新申请。成功入选者将被引导加入生物漏洞赏金平台。
● 保密协议:所有提示、回复、发现及沟通内容均受保密协议约束。
如何申请
请于2025年7月29日前通过此链接提交简短申请(新窗口打开)[https://openai.smapply.org/prog/bio_bug_bounty],内容包括姓名、所属机构、简要经历及150字的计划说明。被接受的申请者及合作伙伴需拥有现有ChatGPT账户,并签署保密协议。
立即申请,助力我们打造更安全的前沿人工智能。
申请OpenAI生物漏洞赏金计划
申请链接:[https://openai.smapply.org/prog/bio_bug_bounty/]
----------------------
Invitation
As part of our ongoing efforts to strengthen our safeguards for advanced AI capabilities in biology, our bio bug bounty is now open for applications. We’ve deployed the ChatGPT agent model and are actively working to further strengthen safety protections for ChatGPT agent and other models. We’re inviting researchers with experience in AI red teaming, security, or chemical and biological risk to try to find a universal jailbreak that can defeat our ten-level bio/chem challenge.
Program overview
● Model in scope: ChatGPT agent only.
● Challenge: Identify one universal jailbreaking prompt to successfully answer all ten bio/chem safety questions from a clean chat.
● Rewards: • $25,000 to the first true universal jailbreak to clear all ten questions.
• $10,000 to the first team that answers all ten questions with multiple jailbreak prompts.
• Smaller awards may be granted for partial wins at our discretion.
● Timeline: Applications open July 17, 2025 with rolling acceptances. Testing begins July 29, 2025.
● Access: Application and invite-only. We will extend invitations to a vetted list of trusted bio red-teamers and review new applications. Once selected, successful applicants will be onboarded to the bio bug bounty platform.
● Disclosure: All prompts, completions, findings, and communications are covered by NDA.
How to apply
Submit a short application here (name, affiliation, brief track record, and a 150-word plan) by July 29, 2025. Accepted applicants and collaborators must have existing ChatGPT accounts to apply, and will sign a NDA.
Apply now and help us make frontier AI safer.
Apply to the OpenAI bio bug bounty
Apply
via OpenAI News