The new agent in ChatGPT from OpenAI has opened up the possibility of performing a wide range of complex tasks directly in the chat, including running code, creating presentations, market analysis, and shopping planning.
This is reported by Business • Media
Innovative capabilities of the AI agent in ChatGPT
OpenAI has announced the launch of a universal agent in ChatGPT that can act on behalf of the user and perform multi-level digital tasks. Key features include creating presentations, analyzing the competitive landscape, automating planning, and organizing purchases. The agent accesses connected applications such as Gmail or GitHub, allowing it to work with real user data.
- The ChatGPT agent is available to subscribers of the Pro, Plus, and Team plans.
- Interaction with external applications is implemented through APIs and terminals, expanding the platform’s capabilities.
- OpenAI has implemented strict memory limitations and additional security measures to prevent biological threats.
Security and unique use cases
The ChatGPT agent combines features from previous developments, such as Deep Research and website navigation, ensuring interaction through natural language prompts and utilizing various tools. The new mode is accessible through a special menu in the service interface.
OpenAI notes that the agent is the first step towards creating a platform capable of not only generating responses but also performing complex actions. For example, the agent can automatically plan a Japanese breakfast or conduct a detailed competitor analysis and prepare a presentation. Such a scenario involves a complete cycle—from data searching and gathering to structuring and managing the process in real-time.
The agent is built on a model that has shown high performance in tests for AI models. At the same time, the company notes that the product has potentially dangerous capabilities, including high “effectiveness” in biological topics.
To mitigate risks, all requests undergo biological classification, and suspicious ones go through additional verification. Furthermore, the memory function has currently been disabled to avoid potential data theft attacks. The possibility of restoring memory will be considered after a thorough risk assessment.
Despite the stated capabilities, it remains unknown how reliably the agent will perform under real user conditions. OpenAI emphasizes that it has developed an even more advanced version, but its effectiveness can only be fully assessed after widespread implementation among users.
Let us remind you that experts from the Future of Life Institute (FLI) emphasize that AI model developers are “absolutely unprepared” for the potential threats that artificial superintelligence may pose.