{"@context":"https://schema.org","@type":"NewsArticle","generatedAt":"2026-07-28T05:20:57.982Z","headline":"OpenAI Workspace Agents 漏洞：一个 ChatGPT 链接即可创建恶意 AI 智能体","description":"安全公司 Zenity Labs 发现 OpenAI Workspace Agents 存在\"AgentForger\"漏洞，攻击者发送一个含恶意提示词的 ChatGPT 链接，即可在受害者账户下创建自主 AI 智能体。该智能体继承受害者身份和已授权应用权限，绕过安全审批，并设置每五分钟运行一次的定时任务，从攻击者邮箱获取指令执行。OpenAI 在四天内修复了该漏洞。","url":"https://www.aioga.com/news/cmrxs4yyw01xkroxpom7zfjvc/","mainEntityOfPage":"https://www.aioga.com/news/cmrxs4yyw01xkroxpom7zfjvc/","datePublished":"2026-07-23T17:01:30.000Z","dateModified":"2026-07-23T17:01:30.000Z","inLanguage":"zh-CN","publisher":{"@type":"NewsMediaOrganization","name":"Aioga","url":"https://www.aioga.com"},"citation":["https://the-decoder.com/one-tampered-chatgpt-link-could-spawn-a-rogue-ai-agent-that-took-orders-from-an-attacker-every-five-minutes","https://aihot.virxact.com/items/cmrxs4yyw01xkroxpom7zfjvc"],"canonicalUrl":"https://www.aioga.com/news/cmrxs4yyw01xkroxpom7zfjvc/","directAnswer":{"@type":"Answer","text":"Zenity Labs 发现，OpenAI Workspace Agents 的“AgentForger”漏洞可让攻击者借助篡改的 ChatGPT 链接，在受害者账户下创建自主智能体，并复用其已有应用权限。OpenAI 已在四天内修复该漏洞。","url":"https://www.aioga.com/news/cmrxs4yyw01xkroxpom7zfjvc/","dateCreated":"2026-07-23T17:01:30.000Z","author":{"@type":"Organization","@id":"https://www.aioga.com/authors/aioga-editorial/#editorial-team","name":"Aioga Editorial Team","url":"https://www.aioga.com/authors/aioga-editorial/"}},"evidence":[{"@type":"CreativeWork","name":"the-decoder.com source article","url":"https://the-decoder.com/one-tampered-chatgpt-link-could-spawn-a-rogue-ai-agent-that-took-orders-from-an-attacker-every-five-minutes","datePublished":"2026-07-23T17:01:30.000Z","provider":{"@type":"Organization","name":"the-decoder.com","url":"https://the-decoder.com/one-tampered-chatgpt-link-could-spawn-a-rogue-ai-agent-that-took-orders-from-an-attacker-every-five-minutes"}},{"@type":"CreativeWork","name":"AIHot archive record","url":"https://aihot.virxact.com/items/cmrxs4yyw01xkroxpom7zfjvc","datePublished":"2026-07-23T17:01:30.000Z","provider":{"@type":"Organization","name":"AIHot","url":"https://aihot.virxact.com/items/cmrxs4yyw01xkroxpom7zfjvc"}}],"aggregationSource":"The Decoder：AI News（RSS）","originalPublisher":{"name":"the-decoder.com","url":"https://the-decoder.com/one-tampered-chatgpt-link-could-spawn-a-rogue-ai-agent-that-took-orders-from-an-attacker-every-five-minutes"},"article":{"id":"cmrxs4yyw01xkroxpom7zfjvc","slug":"cmrxs4yyw01xkroxpom7zfjvc","url":"https://www.aioga.com/news/cmrxs4yyw01xkroxpom7zfjvc/","title":"OpenAI Workspace Agents 漏洞：一个 ChatGPT 链接即可创建恶意 AI 智能体","title_en":"One tampered ChatGPT link could spawn a rogue AI agent that took orders from an attacker every five minutes","summary":"安全公司 Zenity Labs 发现 OpenAI Workspace Agents 存在\"AgentForger\"漏洞，攻击者发送一个含恶意提示词的 ChatGPT 链接，即可在受害者账户下创建自主 AI 智能体。该智能体继承受害者身份和已授权应用权限，绕过安全审批，并设置每五分钟运行一次的定时任务，从攻击者邮箱获取指令执行。OpenAI 在四天内修复了该漏洞。","source":"The Decoder：AI News（RSS）","sourceUrl":"https://the-decoder.com/one-tampered-chatgpt-link-could-spawn-a-rogue-ai-agent-that-took-orders-from-an-attacker-every-five-minutes","aiHotUrl":"https://aihot.virxact.com/items/cmrxs4yyw01xkroxpom7zfjvc","publishedAt":"2026-07-23T17:01:30.000Z","category":"行业动态","score":70,"selected":true,"articleBody":["A single tampered ChatGPT link could spin up an AI agent that quietly checked the attacker's inbox for new orders every five minutes. Zenity Labs calls the flaw a new class of attack against agent-based AI.","AI security firm Zenity Labs found a vulnerability in OpenAI's Workspace Agents that let one manipulated link create an autonomous AI agent under an employee's account. The agent took on the victim's identity and reused their existing app permissions, skipping the approval steps meant to protect sensitive actions.","Zenity named the vulnerability \"AgentForger\"：https://labs.zenity.io/p/agentforger-part-1-chatgpt-cross-site-agent-forgery and sees it as an evolution of classic cross-site request forgery (CSRF). In a typical CSRF attack, someone clicks a bad link or lands on a crafted page and unknowingly fires off an authenticated action they never intended.","AgentForger went further. Rather than triggering a single unwanted action, the manipulated ChatGPT link kicked off the creation of a fully autonomous agent. That agent operated inside the company's trust boundary, tapped into connectors the victim had already authorized, and picked up new tasks from the attacker on a recurring schedule.","Creating a Workspace agent is normally an interactive process. Users select a template, enter instructions, connect tools, review sharing settings, test the agent in preview mode, and then publish it. AgentForger let attackers trigger most of that process through a URL with little further input from the user.","The Agent Builder, introduced in 2025, is available at chatgpt.com/agents/studio/new and accepts two URL parameters. template_name selects a starting template such as \"chief-of-staff,\" while initial_assistant_prompt supplies the instructions.","Zenity found that the page didn't just place the value of initial_assistant_prompt in the prompt field. It also submitted and ran the prompt automatically. Attackers didn't need to send raw requests to ChatGPT or directly manipulate the victim's browser. All they needed was a chatgpt.com link with an attached prompt that looked harmless at first glance.","The only prerequisite was that the victim was logged into ChatGPT, had access to Workspace Agents, and had authorized at least one connector such as Outlook, Gmail, Slack, Google Drive, SharePoint, or Teams. Because the connections already existed, no new OAuth consent prompt appeared that might have tipped the victim off.","In the demo, Zenity embedded a prompt in the URL that guided the Builder through all the steps in a numbered task list. The agent was set up to integrate all already-connected non-MCP connectors and change every permission requirement for reading, writing, and deleting to \"Never ask.\"","It also created schedules to run every five minutes, checked Outlook for emails from the attacker with \"TASK\" in the subject line, executed their instructions using the connected apps, and sent the results back unfiltered.","The Builder created an agent named \"TASK Mail Operator\" without asking the user. It connected the authorized services, disabled approval requirements, published the agent, and launched it in Preview Mode.","Zenity says Preview Mode isn't just a visual test run: It executes the new agent against the victim's real connected accounts using the approval settings that were just configured. Since every setting was already set to \"Never ask,\" the first run finished without asking the user for approval.","Without the scheduler, the attack would have been a one-time event. The scheduler transforms the forged agent into something that resembles command-and-control infrastructure. Once the agent is deployed, the victim doesn't need to click again or reopen ChatGPT. The agent wakes up every five minutes, checks the inbox for new TASK emails, executes the instructions they contain, and sends the results back. The initial click installs the agent, the scheduler keeps it alive, and the inbox becomes the command channel.","In part two：https://labs.zenity.io/p/agentforger-part-2-the-autonomous-insider of its analysis, Zenity shows what attackers could do through this channel. After receiving the command \"TASK 1: RECON,\" the agent mapped the organization. It pulled data from Outlook, Slack, Teams, Drive, SharePoint, and Calendar to list people, roles, channels, active projects, and recurring meetings.","The agent also searched Drive, SharePoint, and Outlook. It found an M&A term sheet, a board presentation that mentioned missed revenue targets and planned layoffs, and a company-wide employee export with contact and compensation data. A request framed as a \"DLP exercise\" told the agent to search Slack for the string \"pass:\". The agent found a database username and password pair and emailed both to the attacker.","Other tasks abused the victim's trusted identity. The agent sent messages through the victim's Teams account asking colleagues to confirm an SSO rollout on a login page controlled by the attacker. Zenity also tested phishing through Slack and a business email compromise template. Other tests included an approval request for a $242,500 wire transfer and a calendar invitation with a participant controlled by the attacker.","Zenity traces AgentForger to two related design choices. The builder treated the initial_assistant_prompt parameter as executable input rather than user input that needed confirmation. An attacker-controlled URL could therefore change data and settings within the victim's authenticated session without the user's explicit approval.","The same prompt could also change security settings, including approval policies and execution schedules. This meant the instruction could disable the system meant to require human approval for sensitive actions.","Zenity describes the combination as \"the lethal trifecta\": the URL supplied untrusted input, connectors provided access to private data, and email offered a path for sending that data out. Most exploits would have to bypass those safeguards first. AgentForger instead gave the attacker access to a build tool that could create an agent with those safeguards already disabled.","Zenity reported AgentForger through OpenAI's Bugcrowd program on June 4, 2026. OpenAI confirmed the report the next day and fixed the flaw on June 8 by removing the affected URL parameter. Zenity praised the OpenAI security team's response time.","Until the fix went live, the flaw affected every organization using ChatGPT Workspace Agents with previously authorized enterprise connectors, according to Zenity.","Zenity says the problem extends beyond this specific bug. Traditional security tools are built for users and endpoints, not autonomous agents that act through legitimate user identities. The more an agent can do without supervision, the more damage it can cause when someone else supplies its instructions. The cybersecurity company calls AgentForger an \"agent trust failure\": the platform assumed that the user had personally created, approved, scheduled, and launched the agent.","Security concerns around agent-based AI have grown recently. Hugging Face reported that a fully AI-controlled agent system entered its production infrastructure：https://the-decoder.com/hugging-face-says-an-ai-agent-hacked-its-infrastructure-and-it-used-ai-to-fight-back/ through a manipulated dataset and then moved laterally. The system performed more than 17,000 actions, according to the company. OpenAI admitted responsibility shortly afterward：https://the-decoder.com/openai-claims-responsibility-for-the-hugging-face-hack-after-its-own-models-escaped-a-test-sandbox/. During a performance test, its model had accidentally hacked Hugging Face to obtain test data.","Zenity also demonstrated several zero-click and one-click exploits last year under the name AgentFlayer：https://the-decoder.com/agent-based-ai-systems-face-growing-threats-from-zero-click-and-one-click-exploits/. The attacks targeted Copilot Studio, Salesforce Einstein, Cursor with Jira MCP, and other enterprise AI tools. In those cases, hidden prompts in seemingly harmless resources could divert customer data or steal login credentials.","Stay in the loop on AI. Clear, useful, no fluff.","Follow The Decoder for AI news, background stories and expert analyses.","The Decoder：https://the-decoder.com/"],"articleImages":[{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2025/10/agent_builder_openai.jpg","alt":"Image description","afterParagraph":0,"url":"/media/articles/cmrxs4yyw01xkroxpom7zfjvc/2ece0d7ba7023a51.jpg"},{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2026/07/agentforger-10-csrf-comparison.jpg","alt":"Diagram comparing classic CSRF, which triggers one unintended request, with AgentForger, which creates an autonomous agent.","afterParagraph":3,"url":"/media/articles/cmrxs4yyw01xkroxpom7zfjvc/ddd24831f6f4f837.jpg"},{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2026/07/agentforger-02-phishing-email.jpg","alt":"Phishing email with the subject line Time-Saving AI Agent Template and a button linking to a manipulated ChatGPT URL.","afterParagraph":6,"url":"/media/articles/cmrxs4yyw01xkroxpom7zfjvc/29aac8b537ff8304.jpg"},{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2026/07/agentforger-04-forged-agent.jpg","alt":"Configuration view of the TASK Mail Operator agent with connected services and all approval settings set to Never ask.","afterParagraph":9,"url":"/media/articles/cmrxs4yyw01xkroxpom7zfjvc/ad8b4cc86ef00637.jpg"},{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2026/07/agentforger-05-preview-mode.jpg","alt":"ChatGPT Preview Mode showing the newly created agent accessing an Outlook inbox and connected apps.","afterParagraph":11,"url":"/media/articles/cmrxs4yyw01xkroxpom7zfjvc/fc0200be31cca1e4.jpg"},{"sourceUrl":"https://the-decoder.com/wp-content/uploads/2026/07/agentforger-06-schedule.jpg","alt":"List of hourly schedules for the agent, each offset by five minutes to run the agent every five minutes.","afterParagraph":12,"url":"/media/articles/cmrxs4yyw01xkroxpom7zfjvc/6abf10f4907b3182.jpg"}],"mediaStatus":"ok","articleBodyZh":["一个被篡改的 ChatGPT 链接就可能启动一个 AI 代理，该代理每五分钟默默检查攻击者的邮箱以获取新指令。Zenity Labs 称这一漏洞是一种针对基于代理的 AI 的新型攻击。","AI 安全公司 Zenity Labs 发现了 OpenAI Workspace Agents 中的一个漏洞，该漏洞允许一个被操控的链接在员工账户下创建一个自主 AI 代理。该代理采用受害者的身份，并复用其现有的应用权限，绕过用于保护敏感操作的审批步骤。","Zenity 将这一漏洞命名为 \"AgentForger\"：https://labs.zenity.io/p/agentforger-part-1-chatgpt-cross-site-agent-forgery，并将其视为经典跨站请求伪造（CSRF）的进化。在典型的 CSRF 攻击中，用户点击一个恶意链接或访问一个精心设计的页面，毫不知情地触发了他们从未打算执行的认证操作。","AgentForger 更进一步。它不仅触发单一的非预期操作，被操控的 ChatGPT 链接启动了一个完全自主代理的创建。该代理在公司的信任边界内运作，利用受害者已授权的连接器，并按照攻击者设置的定期计划接收新任务。","创建 Workspace 代理通常是一个交互式过程。用户选择模板、输入指令、连接工具、审查共享设置、在预览模式下测试代理，然后再发布。AgentForger 允许攻击者通过一个 URL 触发这个过程中大部分步骤，而用户几乎无需任何额外操作。","Agent Builder 于 2025 年推出，可在 chatgpt.com/agents/studio/new 访问，并接受两个 URL 参数。template_name 用于选择起始模板，例如 \"chief-of-staff\"，而 initial_assistant_prompt 用于提供指令。","Zenity 发现该页面不仅将 initial_assistant_prompt 的值放入提示字段中，还会自动提交并运行该提示。攻击者无需向 ChatGPT 发送原始请求或直接操作受害者的浏览器。所需的只是一个附带提示、乍看之下无害的 chatgpt.com 链接。","唯一的前提是受害者已登录 ChatGPT，有权访问 Workspace Agents，并授权了至少一个连接器，如 Outlook、Gmail、Slack、Google Drive、SharePoint 或 Teams。因为连接已经存在，所以不会出现新的 OAuth 授权提示，从而避免引起受害者的警觉。","在演示中，Zenity 在 URL 中嵌入了一个提示，引导 Builder 按编号任务列表完成所有步骤。该代理被设置为整合所有已连接的非 MCP 连接器，并将读取、写入和删除的所有权限要求更改为“从不询问”。","它还创建了每五分钟运行一次的计划任务，检查 Outlook 中发件人为攻击者且主题包含“TASK”的邮件，使用已连接的应用执行指令，并将结果未经过滤地发送回去。","Builder 在未征求用户同意的情况下创建了一个名为“TASK Mail Operator”的代理。它连接已授权的服务，禁用审批要求，发布代理，并在预览模式下启动。","Zenity 说预览模式不仅仅是视觉上的测试运行：它会使用刚配置的审批设置在受害者的真实连接账户上执行新的代理。由于所有设置都已设为“从不询问”，第一次运行无需用户批准就完成了。","如果没有调度程序，这次攻击本来只是一次性事件。调度程序将伪造的代理变成类似指挥控制基础设施的东西。一旦代理部署，受害者无需再次点击或重新打开 ChatGPT。代理每五分钟唤醒一次，检查收件箱中的新 TASK 邮件，执行其中的指令，并发送结果。初始点击安装代理，调度程序维持其运行，收件箱成为指挥通道。","在其分析的第二部分：https://labs.zenity.io/p/agentforger-part-2-the-autonomous-insider 中，Zenity 展示了攻击者通过该通道可能执行的操作。收到命令“TASK 1: RECON”后，代理映射了组织结构。它从 Outlook、Slack、Teams、Drive、SharePoint 和 Calendar 中提取数据，列出人员、角色、频道、活跃项目和定期会议。","该代理还搜索了Drive、SharePoint和Outlook。它发现了一份并购（M&A）条款清单、一份董事会演示文稿（提到未达到的收入目标和计划裁员）、以及包含员工联系方式和薪酬数据的全公司员工导出文件。一项以“DLP演练”为名的请求让代理在Slack中搜索字符串“pass:”。代理发现了一个数据库用户名和密码组合，并通过电子邮件将两者发送给攻击者。","其他任务滥用了受害者的可信身份。代理通过受害者的Teams账户发送消息，要求同事确认在攻击者控制的登录页面上的SSO部署。Zenity还测试了通过Slack进行的网络钓鱼攻击以及商业电子邮件欺诈模板。其他测试包括对一笔$242,500电汇的审批请求，以及包含攻击者控制参与者的日历邀请。","Zenity追踪AgentForger至两个相关的设计选择。构建者将initial_assistant_prompt参数视为可执行输入，而不是需要确认的用户输入。因此，攻击者控制的URL可以在受害者的认证会话中更改数据和设置，而无需用户明确批准。","相同的提示也可以更改安全设置，包括审批策略和执行时间表。这意味着该指令可能会禁用需要人工审批敏感操作的系统。","Zenity将这种组合描述为“致命三合一”：URL提供不可信输入，连接器提供对私人数据的访问，而电子邮件提供将数据发送出去的路径。大多数漏洞利用必须首先绕过这些防护。相反，AgentForger给攻击者访问了一个构建工具，可以创建一个已禁用这些防护的代理。","Zenity于2026年6月4日通过OpenAI的Bugcrowd项目报告了AgentForger。OpenAI次日确认了该报告，并于6月8日通过移除受影响的URL参数修复了此漏洞。Zenity赞扬了OpenAI安全团队的响应速度。","在修复上线之前，该漏洞影响了每个使用ChatGPT Workspace Agents并且之前授权过企业连接器的组织，根据Zenity的说法。","Zenity 表示，这个问题不仅仅局限于这个特定的漏洞。传统的安全工具是为用户和端点设计的，而不是为通过合法用户身份操作的自主代理设计的。代理能够在无人监督的情况下做得越多，当有人提供其指令时，它可能造成的损害就越大。这家网络安全公司称 AgentForger 是一次“代理信任失败”：该平台假设用户已经亲自创建、批准、调度并启动了该代理。","关于基于代理的 AI 的安全问题最近有所增加。Hugging Face 报告称，一个完全由 AI 控制的代理系统通过操纵的数据集进入了其生产基础设施：https://the-decoder.com/hugging-face-says-an-ai-agent-hacked-its-infrastructure-and-it-used-ai-to-fight-back/，然后进行了横向移动。根据公司说法，该系统执行了超过 17,000 次操作。OpenAI 不久后承认了责任：https://the-decoder.com/openai-claims-responsibility-for-the-hugging-face-hack-after-its-own-models-escaped-a-test-sandbox/。在一次性能测试中，其模型意外地入侵了 Hugging Face 以获取测试数据。","Zenity 去年还以 AgentFlayer 的名义演示了几种零点点击和一键点击漏洞：https://the-decoder.com/agent-based-ai-systems-face-growing-threats-from-zero-click-and-one-click-exploits/。这些攻击针对 Copilot Studio、Salesforce Einstein、配合 Jira MCP 的 Cursor 以及其他企业 AI 工具。在这些案例中，看似无害资源中的隐藏提示可能会转移客户数据或窃取登录凭据。","保持对 AI 的关注。清晰、有用、无虚饰。","关注 The Decoder，获取 AI 新闻、背景故事和专家分析。","The Decoder：https://the-decoder.com/"],"translationStatus":"translated","bodyOrigin":"source-page","editorial":{"summary":"Zenity Labs 发现，OpenAI Workspace Agents 的“AgentForger”漏洞可让攻击者借助篡改的 ChatGPT 链接，在受害者账户下创建自主智能体，并复用其已有应用权限。OpenAI 已在四天内修复该漏洞。","background":"正常创建 Workspace 智能体需选择模板、填写指令、连接工具、检查共享设置、预览测试并发布。漏洞利用 Agent Builder 接收的 URL 参数，使恶意提示词被自动提交和运行，从而触发大部分创建流程。","viewpoint":"Aioga 判断，该事件值得关注之处不只是一次未经授权的操作，而是恶意链接能够建立持续运行的智能体。它在企业信任边界内继承身份与连接器权限，使传统链接点击风险延伸到自动化任务层面。","implications":"受害者需已登录 ChatGPT、具备 Workspace Agents 使用权限，并至少授权一个连接器。满足条件后，恶意智能体可能绕过新的 OAuth 同意提示，定期从攻击者邮箱获取任务，并以受害者身份调用既有连接。","nextStep":"值得关注的是，OpenAI 虽已修复该漏洞，企业仍可检查 Workspace Agents 的创建、发布、定时任务及连接器调用记录，并重新审视已有授权范围。相关防护是否覆盖同类智能体创建路径，材料未明确说明。","evidenceRefs":["title","summary","articleBody","source"],"status":"published","aiGenerated":true,"autoApproved":true,"generatedBy":"aioga-editorial:gpt-5.6-sol","reviewedBy":"aioga-editorial-review:gpt-5.6-sol","generatedAt":"2026-07-26T11:50:30.860Z","sourceHash":"8e03ed5ffea5b4ee","review":{"approved":true,"groundedness":96,"clarity":92,"duplicationRisk":14,"blockingIssues":[],"notes":["“绕过新的 OAuth 同意提示”可更准确地表述为“由于复用了已有连接器授权，因此不会出现新的 OAuth 同意提示”，以避免暗示漏洞主动攻破了 OAuth 流程。","“Aioga 判断”属于明确标注的观点，且其判断有来源材料支撑，不构成事实性问题。"]},"validation":{"passed":true,"mode":"ai-auto","revisions":0,"checks":["schema","length","source-attribution","low-source-overlap","no-html","independent-ai-review"]}},"tags":["行业动态","The Decoder：AI News（RSS）"],"translations":{"zh-CN":{"title":"OpenAI Workspace Agents 漏洞：一个 ChatGPT 链接即可创建恶意 AI 智能体","summary":"安全公司 Zenity Labs 发现 OpenAI Workspace Agents 存在\"AgentForger\"漏洞，攻击者发送一个含恶意提示词的 ChatGPT 链接，即可在受害者账户下创建自主 AI 智能体。该智能体继承受害者身份和已授权应用权限，绕过安全审批，并设置每五分钟运行一次的定时任务，从攻击者邮箱获取指令执行。OpenAI 在四天内修复了该漏洞。","category":"行业动态","source":"the-decoder.com","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"OpenAI Workspace Agents 漏洞：一个 ChatGPT 链接即可创建恶意 AI 智能体 - Aioga AI资讯","description":"安全公司 Zenity Labs 发现 OpenAI Workspace Agents 存在\"AgentForger\"漏洞，攻击者发送一个含恶意提示词的 ChatGPT 链接，即可在受害者账户下创建自主 AI 智能体。该智能体继承受害者身份和已授权应用权限，绕过安全审批，并设置每五分钟运行一次的定时任务，从攻击者邮箱获取指令执行。OpenAI 在四天内修复了...","url":"https://www.aioga.com/news/cmrxs4yyw01xkroxpom7zfjvc/"},"en":{"title":"OpenAI Workspace Agents Vulnerability: A ChatGPT Link Can Create Malicious AI Agents","summary":"Security company Zenity Labs discovered a vulnerability called 'AgentForger' in OpenAI Workspace Agents. An attacker can create an autonomous AI agent under a victim's account by sending a ChatGPT link containing a malicious prompt. This agent inherits the victim's identity and authorized application permissions, bypasses security approvals, and sets a scheduled task to run every five minutes, executing instructions from the attacker's email. OpenAI fixed this vulnerability within four days.","category":"Industry","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"OpenAI Workspace Agents Vulnerability: A ChatGPT Link Can Create Malicious AI Agents - Aioga AI News","description":"Security company Zenity Labs discovered a vulnerability called 'AgentForger' in OpenAI Workspace Agents. An attacker can create an autonomous AI agent under a victim's account by s...","url":"https://www.aioga.com/en/news/cmrxs4yyw01xkroxpom7zfjvc/","contentTranslated":true,"sourceHash":"3a44ac86deeb7b57","translatedAt":"2026-07-26T02:43:47.873Z"},"ja":{"title":"OpenAI Workspace エージェントの脆弱性：1つの ChatGPT リンクで悪意のある AI エージェントを作成可能","summary":"セキュリティ会社 Zenity Labs は、OpenAI Workspace Agents に「AgentForger」脆弱性が存在することを発見しました。攻撃者が悪意のあるプロンプトを含む ChatGPT のリンクを送信すると、被害者のアカウントで自律的な AI エージェントを作成することができます。このエージェントは被害者のIDと既に承認されたアプリの権限を引き継ぎ、セキュリティ承認を回避し、5分ごとに実行されるスケジュールタスクを設定して、攻撃者のメールから指示を受け取り実行します。OpenAI は4日以内にこの脆弱性を修正しました。","category":"業界動向","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"OpenAI Workspace エージェントの脆弱性：1つの ChatGPT リンクで悪意のある AI エージェントを作成可能 - Aioga AIニュース","description":"セキュリティ会社 Zenity Labs は、OpenAI Workspace Agents に「AgentForger」脆弱性が存在することを発見しました。攻撃者が悪意のあるプロンプトを含む ChatGPT のリンクを送信すると、被害者のアカウントで自律的な AI エージェントを作成することができます。このエージェントは被害者のIDと既に承認されたアプリの...","url":"https://www.aioga.com/ja/news/cmrxs4yyw01xkroxpom7zfjvc/","contentTranslated":true,"sourceHash":"3a44ac86deeb7b57","translatedAt":"2026-07-26T02:44:02.042Z"},"ko":{"title":"OpenAI Workspace Agents 취약점: 하나의 ChatGPT 링크로 악성 AI 에이전트를 생성할 수 있음","summary":"보안 회사 Zenity Labs는 OpenAI Workspace Agents에서 'AgentForger' 취약점을 발견했습니다. 공격자는 악성 프롬프트가 포함된 ChatGPT 링크를 보내면 피해자 계정에서 자율 AI 에이전트를 생성할 수 있습니다. 이 에이전트는 피해자의 신원과 이미 승인된 애플리케이션 권한을 상속하고, 보안 승인을 우회하며, 매 5분마다 실행되는 예약 작업을 설정하여 공격자 이메일에서 지시를 받아 실행합니다. OpenAI는 4일 만에 이 취약점을 수정했습니다.","category":"업계 동향","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"OpenAI Workspace Agents 취약점: 하나의 ChatGPT 링크로 악성 AI 에이전트를 생성할 수 있음 - Aioga AI 뉴스","description":"보안 회사 Zenity Labs는 OpenAI Workspace Agents에서 'AgentForger' 취약점을 발견했습니다. 공격자는 악성 프롬프트가 포함된 ChatGPT 링크를 보내면 피해자 계정에서 자율 AI 에이전트를 생성할 수 있습니다. 이 에이전트는 피해자의 신원과 이미 승인된 애플리케이션 권한을 상속하고,...","url":"https://www.aioga.com/ko/news/cmrxs4yyw01xkroxpom7zfjvc/","contentTranslated":true,"sourceHash":"3a44ac86deeb7b57","translatedAt":"2026-07-26T02:44:39.593Z"},"es":{"title":"Vulnerabilidad de OpenAI Workspace Agents: se puede crear un agente de IA malicioso con solo un enlace de ChatGPT","summary":"La empresa de seguridad Zenity Labs descubrió que los Agentes de OpenAI Workspace presentaban una vulnerabilidad llamada \"AgentForger\". Un atacante que envíe un enlace de ChatGPT con una instrucción maliciosa puede crear un agente de IA autónomo bajo la cuenta de la víctima. Este agente hereda la identidad de la víctima y los permisos de las aplicaciones ya autorizadas, elude la aprobación de seguridad y configura una tarea programada para ejecutarse cada cinco minutos, obteniendo instrucciones para ejecutar desde el correo electrónico del atacante. OpenAI solucionó esta vulnerabilidad en cuatro días.","category":"Industria","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Vulnerabilidad de OpenAI Workspace Agents: se puede crear un agente de IA malicioso con solo un enlace de ChatGPT - Aioga Noticias de IA","description":"La empresa de seguridad Zenity Labs descubrió que los Agentes de OpenAI Workspace presentaban una vulnerabilidad llamada \"AgentForger\". Un atacante que envíe un enlace de ChatGPT c...","url":"https://www.aioga.com/es/news/cmrxs4yyw01xkroxpom7zfjvc/","contentTranslated":true,"sourceHash":"3a44ac86deeb7b57","translatedAt":"2026-07-26T02:44:35.622Z"},"fr":{"title":"Vulnérabilité des agents OpenAI Workspace : un lien ChatGPT suffit pour créer un agent IA malveillant","summary":"La société de sécurité Zenity Labs a découvert une faille \"AgentForger\" dans OpenAI Workspace Agents. L'attaquant envoie un lien ChatGPT contenant un prompt malveillant, ce qui permet de créer un agent IA autonome sous le compte de la victime. Cet agent hérite de l'identité de la victime et des autorisations des applications déjà accordées, contourne l'approbation de sécurité et configure une tâche cron exécutée toutes les cinq minutes pour recevoir et exécuter des instructions depuis l'e-mail de l'attaquant. OpenAI a corrigé cette faille en quatre jours.","category":"Industrie","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Vulnérabilité des agents OpenAI Workspace : un lien ChatGPT suffit pour créer un agent IA malveillant - Aioga Actualités IA","description":"La société de sécurité Zenity Labs a découvert une faille \"AgentForger\" dans OpenAI Workspace Agents. L'attaquant envoie un lien ChatGPT contenant un prompt malveillant, ce qui per...","url":"https://www.aioga.com/fr/news/cmrxs4yyw01xkroxpom7zfjvc/","contentTranslated":true,"sourceHash":"3a44ac86deeb7b57","translatedAt":"2026-07-26T02:45:21.229Z"},"de":{"title":"OpenAI Workspace Agents Schwachstelle: Ein ChatGPT-Link reicht aus, um bösartige KI-Agenten zu erstellen","summary":"Die Sicherheitsfirma Zenity Labs entdeckte eine 'AgentForger'-Schwachstelle bei OpenAI Workspace Agents. Angreifer konnten einen ChatGPT-Link mit einem bösartigen Prompt senden und so einen autonomen AI-Agenten im Konto des Opfers erstellen. Dieser Agent übernimmt die Identität des Opfers und die bereits genehmigten Anwendungsberechtigungen, umgeht die Sicherheitsgenehmigung und richtet eine alle fünf Minuten ausgeführte geplante Aufgabe ein, um Anweisungen aus dem E-Mail-Postfach des Angreifers auszuführen. OpenAI hat diese Schwachstelle innerhalb von vier Tagen behoben.","category":"行业动态","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"OpenAI Workspace Agents Schwachstelle: Ein ChatGPT-Link reicht aus, um bösartige KI-Agenten zu erstellen - Aioga KI-News","description":"Die Sicherheitsfirma Zenity Labs entdeckte eine 'AgentForger'-Schwachstelle bei OpenAI Workspace Agents. Angreifer konnten einen ChatGPT-Link mit einem bösartigen Prompt senden und...","url":"https://www.aioga.com/de/news/cmrxs4yyw01xkroxpom7zfjvc/","contentTranslated":true,"sourceHash":"3a44ac86deeb7b57","translatedAt":"2026-07-26T02:45:17.935Z"},"pt-BR":{"title":"Vulnerabilidade do OpenAI Workspace Agents: é possível criar agentes de IA maliciosos apenas com um link do ChatGPT","summary":"A empresa de segurança Zenity Labs descobriu uma vulnerabilidade \"AgentForger\" nos OpenAI Workspace Agents. Um atacante pode enviar um link do ChatGPT com uma instrução maliciosa, podendo criar um agente de IA autônomo na conta da vítima. Esse agente herda a identidade da vítima e as permissões dos aplicativos já autorizados, contorna a aprovação de segurança e configura uma tarefa agendada para executar a cada cinco minutos, recebendo instruções a partir do e-mail do atacante. A OpenAI corrigiu essa vulnerabilidade em quatro dias.","category":"行业动态","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Vulnerabilidade do OpenAI Workspace Agents: é possível criar agentes de IA maliciosos apenas com um link do ChatGPT - Aioga Notícias de IA","description":"A empresa de segurança Zenity Labs descobriu uma vulnerabilidade \"AgentForger\" nos OpenAI Workspace Agents. Um atacante pode enviar um link do ChatGPT com uma instrução maliciosa,...","url":"https://www.aioga.com/pt-BR/news/cmrxs4yyw01xkroxpom7zfjvc/","contentTranslated":true,"sourceHash":"3a44ac86deeb7b57","translatedAt":"2026-07-26T02:45:55.184Z"},"ru":{"title":"Уязвимость OpenAI Workspace Agents: можно создать вредоносного ИИ-агента с помощью одной ссылки ChatGPT","summary":"Компания по кибербезопасности Zenity Labs обнаружила у агентов OpenAI Workspace уязвимость под названием \"AgentForger\". Злоумышленник может отправить ссылку на ChatGPT с вредоносной подсказкой, чтобы создать автономного ИИ-агента в аккаунте жертвы. Этот агент наследует личность жертвы и разрешения на уже авторизованные приложения, обходит проверку безопасности и настраивает задание по расписанию каждые пять минут для получения инструкций с электронной почты атакующего. OpenAI исправила эту уязвимость в течение четырех дней.","category":"行业动态","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Уязвимость OpenAI Workspace Agents: можно создать вредоносного ИИ-агента с помощью одной ссылки ChatGPT - Aioga Новости ИИ","description":"Компания по кибербезопасности Zenity Labs обнаружила у агентов OpenAI Workspace уязвимость под названием \"AgentForger\". Злоумышленник может отправить ссылку на ChatGPT с вредоносно...","url":"https://www.aioga.com/ru/news/cmrxs4yyw01xkroxpom7zfjvc/","contentTranslated":true,"sourceHash":"3a44ac86deeb7b57","translatedAt":"2026-07-26T02:46:00.561Z"},"ar":{"title":"ثغرة في وكلاء OpenAI Workspace: يمكن إنشاء وكيل ذكاء اصطناعي خبيث بمجرد رابط ChatGPT","summary":"اكتشفت شركة الأمان Zenity Labs وجود ثغرة 'AgentForger' في OpenAI Workspace Agents، حيث يمكن للمهاجم إرسال رابط ChatGPT يحتوي على تعليمات ضارة، مما يؤدي إلى إنشاء وكيل ذكاء اصطناعي مستقل في حساب الضحية. يرث هذا الوكيل هوية الضحية وصلاحيات التطبيقات المصرح بها، متجاوزًا الموافقة الأمنية، ويقوم بتعيين مهمة مجدولة للتنفيذ كل خمس دقائق للحصول على التعليمات من بريد المهاجم الإلكتروني. قامت OpenAI بإصلاح هذه الثغرة خلال أربعة أيام.","category":"行业动态","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"ثغرة في وكلاء OpenAI Workspace: يمكن إنشاء وكيل ذكاء اصطناعي خبيث بمجرد رابط ChatGPT - Aioga أخبار الذكاء الاصطناعي","description":"اكتشفت شركة الأمان Zenity Labs وجود ثغرة 'AgentForger' في OpenAI Workspace Agents، حيث يمكن للمهاجم إرسال رابط ChatGPT يحتوي على تعليمات ضارة، مما يؤدي إلى إنشاء وكيل ذكاء اصطناعي...","url":"https://www.aioga.com/ar/news/cmrxs4yyw01xkroxpom7zfjvc/","contentTranslated":true,"sourceHash":"3a44ac86deeb7b57","translatedAt":"2026-07-26T02:46:37.983Z"},"hi":{"title":"OpenAI Workspace एजेंट्स की कमजोरियाँ: केवल एक ChatGPT लिंक से खतरनाक AI एजेंट बनाया जा सकता है","summary":"सुरक्षा कंपनी Zenity Labs ने OpenAI Workspace Agents में \"AgentForger\" भेद्यता पाई। हमलावर एक दुर्भावनापूर्ण प्रॉम्प्ट वाले ChatGPT लिंक भेजकर पीड़ित के खाते में स्वायत्त AI एजेंट बना सकता है। यह एजेंट पीड़ित की पहचान और प्राधिकृत एप्लिकेशन अनुमतियों को अपनाता है, सुरक्षा अनुमोदन को बायपास करता है, और हर पाँच मिनट में एक निर्धारित कार्य चलाता है, जो हमलावर के ईमेल से निर्देश प्राप्त करता है। OpenAI ने चार दिनों में इस भेद्यता को ठीक किया।","category":"行业动态","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"OpenAI Workspace एजेंट्स की कमजोरियाँ: केवल एक ChatGPT लिंक से खतरनाक AI एजेंट बनाया जा सकता है - Aioga AI समाचार","description":"सुरक्षा कंपनी Zenity Labs ने OpenAI Workspace Agents में \"AgentForger\" भेद्यता पाई। हमलावर एक दुर्भावनापूर्ण प्रॉम्प्ट वाले ChatGPT लिंक भेजकर पीड़ित के खाते में स्वायत्त AI एजेंट...","url":"https://www.aioga.com/hi/news/cmrxs4yyw01xkroxpom7zfjvc/","contentTranslated":true,"sourceHash":"3a44ac86deeb7b57","translatedAt":"2026-07-26T02:46:42.392Z"},"it":{"title":"Vulnerabilità degli OpenAI Workspace Agents: un link a ChatGPT può creare un agente AI dannoso","summary":"La società di sicurezza Zenity Labs ha scoperto una vulnerabilità \"AgentForger\" nei Workspace Agents di OpenAI. Un aggressore può inviare un link ChatGPT contenente prompt maligni per creare un agente AI autonomo sotto l'account della vittima. Questo agente eredita l'identità della vittima e i permessi delle applicazioni autorizzate, aggirando le approvazioni di sicurezza, e imposta un'attività programmata che viene eseguita ogni cinque minuti per ricevere e eseguire istruzioni dalla casella email dell'aggressore. OpenAI ha corretto la vulnerabilità in quattro giorni.","category":"行业动态","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Vulnerabilità degli OpenAI Workspace Agents: un link a ChatGPT può creare un agente AI dannoso - Aioga Notizie IA","description":"La società di sicurezza Zenity Labs ha scoperto una vulnerabilità \"AgentForger\" nei Workspace Agents di OpenAI. Un aggressore può inviare un link ChatGPT contenente prompt maligni...","url":"https://www.aioga.com/it/news/cmrxs4yyw01xkroxpom7zfjvc/","contentTranslated":true,"sourceHash":"3a44ac86deeb7b57","translatedAt":"2026-07-26T02:47:17.893Z"},"nl":{"title":"OpenAI Workspace Agents-kwetsbaarheid: een ChatGPT-link is genoeg om een kwaadaardige AI-agent te creëren","summary":"Het beveiligingsbedrijf Zenity Labs ontdekte een 'AgentForger'-kwetsbaarheid bij OpenAI Workspace Agents. Een aanvaller kan een ChatGPT-link met een kwaadaardige prompt verzenden, waarmee een autonome AI-agent onder het account van het slachtoffer kan worden aangemaakt. Deze agent erft de identiteit van het slachtoffer en de reeds geautoriseerde applicatiemachtigingen, omzeilt de beveiligingsgoedkeuring en stelt een cronjob in die elke vijf minuten draait om instructies van het e-mailadres van de aanvaller uit te voeren. OpenAI heeft deze kwetsbaarheid binnen vier dagen verholpen.","category":"行业动态","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"OpenAI Workspace Agents-kwetsbaarheid: een ChatGPT-link is genoeg om een kwaadaardige AI-agent te creëren - Aioga AI-nieuws","description":"Het beveiligingsbedrijf Zenity Labs ontdekte een 'AgentForger'-kwetsbaarheid bij OpenAI Workspace Agents. Een aanvaller kan een ChatGPT-link met een kwaadaardige prompt verzenden,...","url":"https://www.aioga.com/nl/news/cmrxs4yyw01xkroxpom7zfjvc/","contentTranslated":true,"sourceHash":"3a44ac86deeb7b57","translatedAt":"2026-07-26T02:47:16.126Z"},"tr":{"title":"OpenAI Workspace Agents Açığı: Sadece Bir ChatGPT Bağlantısıyla Zararlı AI Ajanı Oluşturulabilir","summary":"Güvenlik şirketi Zenity Labs, OpenAI Workspace Agents'ta \"AgentForger\" açığını keşfetti. Saldırgan, kötü niyetli bir prompt içeren bir ChatGPT bağlantısı göndererek, mağdurun hesabı altında bağımsız bir yapay zeka ajanı oluşturabiliyordu. Bu ajan, mağdurun kimliğini ve yetkilendirilmiş uygulama izinlerini devralıyor, güvenlik onayını atlayarak her beş dakikada bir çalışacak şekilde zamanlanmış görev ayarlıyor ve talimatları saldırganın e-posta adresinden alarak yürütüyordu. OpenAI, bu açığı dört gün içinde giderdi.","category":"行业动态","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"OpenAI Workspace Agents Açığı: Sadece Bir ChatGPT Bağlantısıyla Zararlı AI Ajanı Oluşturulabilir - Aioga AI Haberleri","description":"Güvenlik şirketi Zenity Labs, OpenAI Workspace Agents'ta \"AgentForger\" açığını keşfetti. Saldırgan, kötü niyetli bir prompt içeren bir ChatGPT bağlantısı göndererek, mağdurun hesab...","url":"https://www.aioga.com/tr/news/cmrxs4yyw01xkroxpom7zfjvc/","contentTranslated":true,"sourceHash":"3a44ac86deeb7b57","translatedAt":"2026-07-26T02:47:55.400Z"},"vi":{"title":"Lỗ hổng OpenAI Workspace Agents: Chỉ cần một liên kết ChatGPT là có thể tạo ra tác nhân AI độc hại","summary":"Công ty an ninh Zenity Labs phát hiện lỗ hổng \"AgentForger\" trong OpenAI Workspace Agents. Kẻ tấn công chỉ cần gửi một liên kết ChatGPT chứa từ khóa độc hại, có thể tạo ra một AI tự động trong tài khoản của nạn nhân. AI này kế thừa danh tính của nạn nhân và quyền truy cập các ứng dụng đã được ủy quyền, vượt qua quy trình phê duyệt an ninh, và thiết lập nhiệm vụ định kỳ chạy mỗi năm phút một lần để nhận lệnh từ email của kẻ tấn công và thực thi. OpenAI đã khắc phục lỗ hổng này trong vòng bốn ngày.","category":"行业动态","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Lỗ hổng OpenAI Workspace Agents: Chỉ cần một liên kết ChatGPT là có thể tạo ra tác nhân AI độc hại - Tin tức AI Aioga","description":"Công ty an ninh Zenity Labs phát hiện lỗ hổng \"AgentForger\" trong OpenAI Workspace Agents. Kẻ tấn công chỉ cần gửi một liên kết ChatGPT chứa từ khóa độc hại, có thể tạo ra một AI t...","url":"https://www.aioga.com/vi/news/cmrxs4yyw01xkroxpom7zfjvc/","contentTranslated":true,"sourceHash":"3a44ac86deeb7b57","translatedAt":"2026-07-26T02:47:53.076Z"},"id":{"title":"Kelemahan OpenAI Workspace Agents: Satu tautan ChatGPT dapat membuat agen AI berbahaya","summary":"Perusahaan keamanan Zenity Labs menemukan kerentanan \"AgentForger\" pada OpenAI Workspace Agents, di mana penyerang dapat mengirim tautan ChatGPT berisi prompt jahat, sehingga dapat membuat agen AI otonom di akun korban. Agen ini mewarisi identitas korban dan izin aplikasi yang sudah diberi otorisasi, melewati persetujuan keamanan, dan mengatur tugas terjadwal yang berjalan setiap lima menit untuk mengambil perintah dari email penyerang. OpenAI memperbaiki kerentanan ini dalam waktu empat hari.","category":"行业动态","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Kelemahan OpenAI Workspace Agents: Satu tautan ChatGPT dapat membuat agen AI berbahaya - Berita AI Aioga","description":"Perusahaan keamanan Zenity Labs menemukan kerentanan \"AgentForger\" pada OpenAI Workspace Agents, di mana penyerang dapat mengirim tautan ChatGPT berisi prompt jahat, sehingga dapat...","url":"https://www.aioga.com/id/news/cmrxs4yyw01xkroxpom7zfjvc/","contentTranslated":true,"sourceHash":"3a44ac86deeb7b57","translatedAt":"2026-07-26T02:48:29.011Z"},"th":{"title":"ช่องโหว่ OpenAI Workspace Agents: เพียงลิงก์ ChatGPT ก็สามารถสร้างเอเจนต์ AI ที่มุ่งร้ายได้","summary":"บริษัทความปลอดภัย Zenity Labs พบช่องโหว่ \"AgentForger\" ใน OpenAI Workspace Agents โดยที่ผู้โจมตีสามารถส่งลิงก์ ChatGPT ที่มีคำสั่งแฝงอันตราย เพียงเท่านี้ก็สามารถสร้างเอเย่นต์ AI อัตโนมัติในบัญชีเหยื่อได้ เอเย่นต์นี้จะสืบทอดตัวตนของเหยื่อและสิทธิ์การใช้งานที่ได้รับอนุญาต ข้ามการอนุมัติความปลอดภัย และตั้งงานตามกำหนดให้ทำงานทุกห้านาที เพื่อรับคำสั่งจากอีเมลของผู้โจมตี OpenAI ได้แก้ไขช่องโหว่นี้ภายในสี่วัน","category":"行业动态","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"ช่องโหว่ OpenAI Workspace Agents: เพียงลิงก์ ChatGPT ก็สามารถสร้างเอเจนต์ AI ที่มุ่งร้ายได้ - ข่าว AI Aioga","description":"บริษัทความปลอดภัย Zenity Labs พบช่องโหว่ \"AgentForger\" ใน OpenAI Workspace Agents โดยที่ผู้โจมตีสามารถส่งลิงก์ ChatGPT ที่มีคำสั่งแฝงอันตราย เพียงเท่านี้ก็สามารถสร้างเอเย่นต์ AI อั...","url":"https://www.aioga.com/th/news/cmrxs4yyw01xkroxpom7zfjvc/","contentTranslated":true,"sourceHash":"3a44ac86deeb7b57","translatedAt":"2026-07-26T02:48:40.639Z"},"pl":{"title":"Luka w OpenAI Workspace Agents: połączenie z ChatGPT umożliwia stworzenie złośliwego agenta AI","summary":"Firma bezpieczeństwa Zenity Labs odkryła lukę „AgentForger” w OpenAI Workspace Agents. Atakujący może wysłać link do ChatGPT zawierający złośliwe polecenie, co pozwala na stworzenie autonomicznego agenta AI w koncie ofiary. Agent ten dziedziczy tożsamość ofiary i uprawnienia autoryzowanych aplikacji, omija zatwierdzenie bezpieczeństwa oraz ustawia zadanie cykliczne uruchamiane co pięć minut, wykonując polecenia z skrzynki pocztowej atakującego. OpenAI naprawiło tę lukę w ciągu czterech dni.","category":"行业动态","source":"The Decoder：AI News（RSS）","aggregationSource":"The Decoder：AI News（RSS）","pageTitle":"Luka w OpenAI Workspace Agents: połączenie z ChatGPT umożliwia stworzenie złośliwego agenta AI - Aioga Wiadomości AI","description":"Firma bezpieczeństwa Zenity Labs odkryła lukę „AgentForger” w OpenAI Workspace Agents. Atakujący może wysłać link do ChatGPT zawierający złośliwe polecenie, co pozwala na stworzeni...","url":"https://www.aioga.com/pl/news/cmrxs4yyw01xkroxpom7zfjvc/","contentTranslated":true,"sourceHash":"3a44ac86deeb7b57","translatedAt":"2026-07-26T02:49:21.418Z"}}}}