Summary: We request a feature that requires AI to obtain explicit user consent before conducting any web searches. This would enhance privacy, user control, and AI safety, ensuring that external lookups only occur when intended by the user.
Problem Statement: While some users prefer automatic searches for convenience, others prioritize privacy and control over when AI accesses external information. This request addresses the needs of users who want transparency and explicit consent before any web searches are conducted. Currently, AI systems can perform web searches mid-conversation without always prompting the user first. While some users may find this feature convenient, others have significant privacy and security concerns, including:
- Accidental data exposure: AI might search for sensitive topics without user intent.
- Surveillance risks: Authoritarian governments or corporations could monitor AI-triggered searches in real-time, profiling users without their knowledge.
- Manipulation and bias: Automatic lookups may fetch results from sources that skew discussions in unintended ways.
- Loss of user agency: Users should have full control over what gets searched, rather than AI making the decision.
Proposed Solution: Introduce a built-in user consent system that requires AI to explicitly confirm search requests before execution. This would include:
- A “Permission to Search” Prompt: Before performing any web lookup, AI must ask: “Would you like me to look this up?”
- Clear Search Transparency: Display what will be searched before executing the query.
- User-Toggled Search Settings: A setting in AI models that allows users to choose between:
- Always ask before searching (default).
- Manual search only (AI never searches unless told).
- Automatic search (for users who opt in to instant lookups).
- User-Viewable Search History: A log of all web searches initiated via AI, ensuring accountability and transparency.
Benefits: 
Enhances Privacy: Prevents AI from exposing unintended sensitive queries.
Strengthens User Control: Empowers users to choose what is looked up.
Mitigates Surveillance Risks: Prevents AI-triggered searches from being monitored by governments, corporations, or tracking systems that could profile users based on their interactions. Reduces real-time profiling based on AI-triggered searches.
Improves AI Transparency: Ensures users always know what is being searched and why.
Balances Convenience & Security: Users can seamlessly switch between manual, prompt-based, or automatic search modes depending on their privacy preferences and workflow needs. Allows those who prefer automatic searches to opt in, while others can retain control.
Conclusion: For AI to remain a trusted and safe tool, it must operate with full transparency and user consent. By implementing an explicit opt-in mechanism for web searches, we can ensure AI remains a privacy-conscious, user-driven tool rather than an uncontrolled data-gathering mechanism.
We urge OpenAI and other AI developers to prioritize this as a critical safeguard for AI safety and ethical design, while still allowing flexibility for users who prefer automatic searches.