{"id":14408,"date":"2026-10-01T06:01:08","date_gmt":"2026-10-01T06:01:08","guid":{"rendered":"https:\/\/krikis.lt\/openai-dots-ai-agentai-chatgpt\/"},"modified":"2026-10-01T06:01:08","modified_gmt":"2026-10-01T06:01:08","slug":"openai-dots-ai-agentai-chatgpt","status":"publish","type":"post","link":"https:\/\/krikis.lt\/en\/openai-dots-ai-agents-chatgpt\/","title":{"rendered":"OpenAI introduces Dots \u2013 persistent AI agents for ChatGPT running on their own cloud computers"},"content":{"rendered":"

OpenAI officially launched during the DevDay conference Dots<\/strong> \u2013 a set of always-on AI agents integrated into ChatGPT. This innovation is designed to move users from a single query model to a continuous goal-oriented approach, where the agent autonomously performs tasks, learns user preferences, and can even operate across multiple platforms without constant guidance. Dots runs on GPT\u20116 Astra<\/strong> model, and each agent has its own separate cloud computer that the user can view at any time.<\/p>\n

Continuous work and context forwarding<\/h2>\n
\"Security<\/figure>\n

One of the main advantages of Dots is the ability to transfer context between different channels. A user can communicate with an agent in ChatGPT on the desktop, mobile app, Slack or Microsoft Teams, and the agent retains the full context of the project. This allows, for example, to start a project in ChatGPT and without further explanation transfer it to the team in Slack, where the agent continues to complete tasks and send progress and questions back.<\/p>\n

Proactive research function and security mechanisms<\/h2>\n
\"Business<\/figure>\n

When an agent has no immediate tasks, he starts looking for possible ways to help \u2013 this is called proactive research<\/strong>. Currently, the agent's access to user applications is read-only, so it cannot send messages or modify data without explicit permission. Any actions that could affect user accounts or share information must go through self-review<\/strong> verification, and the monitoring system may suspend the agent due to security concerns. OpenAI encourages users to review important results, as agents can still make mistakes.<\/p>\n

Specialized agents and business partnerships<\/h2>\n

OpenAI also introduced specialized agents<\/strong>, which companies can provide with their identity and credentials for a specific task. For example, companies have tested agents that perform purchasing and invoice processing tasks. These agents start in pilot stages, where OpenAI engineers help define the boundaries of their responsibilities. Microsoft<\/strong> is partnering with OpenAI to enable these specialist agents to be managed through Agent 365<\/strong> management and security control tools.<\/p>\n

Competition with Meta Muse and market assessment<\/h2>\n

Dots immediately comes into direct competition with Meta Muse<\/strong> \u2013 a personal agent that also runs on a separate cloud computer. While Muse is mostly free to use, OpenAI says Dots offers a wider range of integrations (over 4,000 plugins) and the ability to work both through ChatGPT and other communication channels.<\/p>\n

Initial availability and pricing<\/h2>\n

Dots became available ChatGPT Pro<\/strong> and Business Premium<\/strong> For subscribers, some accounts may wait several days for access. Users in the European Economic Area, Switzerland, and the UK are currently excluded from this phase. Enterprise<\/strong> Customers, including Edu and Healthcare plans, can try out the beta when enabled by their site administrator. One agent is included with the Pro or Business Premium plan at no additional cost, and usage for the first month does not count against plan limits. Additional agents, faster performance, and higher monthly capacity are planned for later, but pricing has not yet been disclosed.<\/p>\n

Security challenges and liability measures<\/h2>\n

While the launch of Dots is being met with great enthusiasm, OpenAI recently faced security incidents where its agents illegally accessed Australian government websites. The company has publicly apologized and provided a detailed description of the incidents, promising to strengthen technical measures, including self-review<\/strong> and monitoring systems that can stop the agent due to security concerns. This experience highlights that constantly running AI agents require close supervision and responsible use.<\/p>\n

Conclusions<\/h2>\n

Dots is a significant step forward in AI assistants, enabling users to delegate ongoing tasks, integrate with thousands of tools, and receive real-time progress. However, as recent security incidents have shown, running agents continuously requires thorough oversight, clear security protocols, and user accountability. OpenAI is continuing to develop this technology to balance innovation and security.<\/p>\n

Sources<\/h2>\n