OpenAI CEO Sam Altman has outlined a vision for a future version of ChatGPT that continuously monitors your digital activity, including your computer screen, meetings, calls, emails, and documents, to provide proactive assistance without requiring detailed manual prompts. Altman believes this level of context-aware AI could become practically useful within the next six months, representing a fundamental shift in how people interact with artificial intelligence.

How Could an Always-On AI Assistant Transform Your Workday?

The proposed system would fundamentally change the relationship between users and AI. Instead of opening ChatGPT, typing a specific request, and waiting for a response, users could have an AI assistant working alongside them throughout the day, understanding their ongoing activities without being asked. For someone preparing a presentation, the AI could review documents already in use, recall previous discussions, and identify overlooked information. For a business owner, it could help track conversations, unfinished tasks, and important decisions.

• Screen Monitoring: The AI would observe what is happening on a user's computer and understand the context of their work in real time.
• Meeting and Call Integration: The system could transcribe and understand meetings and calls to provide relevant follow-up assistance.
• Communication Access: The assistant would connect with email, text messages, documents, and workplace tools like Slack to build a complete picture of user activity.
• Proactive Problem Detection: An AI with surrounding context could notice inconsistencies in documents or alert users to important follow-ups discussed in meetings but never added to task lists.

What Privacy and Security Challenges Does This Vision Raise?

The potential benefits are significant, but so are the risks. An AI system that can see your screen, understand meetings, and access communications could have extraordinary amounts of information about your life, including private conversations, work documents, financial details, personal plans, and confidential business information. According to reporting on Altman's comments, the proposed model would depend on users deciding which information and services the AI could access, creating a distinction between an assistant that users deliberately allow to access their work environment and one that silently observes everything.

Questions around data retention, security, consent, and deletion could become just as important as the AI's capabilities themselves. OpenAI has already introduced some foundational features pointing toward this vision. The company recently improved ChatGPT's memory to maintain useful information over longer periods and provide greater continuity between interactions. OpenAI also introduced Computer History on its Mac desktop application, which allows ChatGPT to use activity information from selected applications and websites to gain additional context about what users have been doing. Users can control which apps and websites contribute information and can pause or delete collected activity.

Is This Vision Actually Achievable in Six Months?

Altman's six-month timeframe should be understood as a prediction about when this kind of AI could become practically useful, rather than a confirmed announcement that OpenAI will launch a specific "always watching" ChatGPT feature in six months. Current reporting describes his comments as a vision for a descendant of ChatGPT rather than a confirmed product release.

There is also a significant engineering challenge involved in bringing all these capabilities together. Screen understanding, memory, meeting transcription, tool use, and long-term context already exist in different forms across various AI systems. The harder problem is combining them reliably while ensuring that the AI understands what information actually matters and can act on it appropriately.

The most important aspect of Altman's prediction may not be the idea of ChatGPT watching a screen. Instead, it represents a move from prompt-based AI to context-based AI. Today, users generally have to tell an AI what they need. In the future, the assistant may already understand the situation before the user asks. That could make AI dramatically more useful for work, education, and everyday organization. But it also means people will have to decide how much of their lives they are willing to share with an AI system.

The next generation of AI may therefore compete not only on intelligence, but on something much more personal: how much of your life you trust it to understand. This shift represents one of the most significant changes in how AI assistants could integrate into daily life, raising important questions about the balance between convenience and privacy that users will need to navigate.