[claude.ai] Input Field Modifies User Content Without Consent
Problem
The claude.ai input field modifies user-typed content without consent. Numbers and other content are altered after the user types them.
Observed Behavior
- User types a specific number (e.g., "1000")
- The value changes to something else without user action
- Other content alterations have been observed
Expected Behavior
User input is sacred. The interface should never, under any circumstances, modify what the user has typed. What goes in must come out exactly as typed.
Impact
- Users cannot trust that their input will be preserved
- Particularly problematic for:
- Code snippets
- Technical specifications
- Numerical data
- Precise instructions
- Forces users to double-check every input before sending
Proposed Solutions
- Audit all input processing - Identify any transformation, formatting, or "smart" features that modify input
- Remove or make optional any auto-formatting features
- Preserve raw input - Store and send exactly what the user typed
- If transformation is needed for display, keep original in state
Principle
Input fields should be pass-through. The user's keystrokes are their intent. Modifying them is a violation of basic UX principles.
---
Alternatively, if these fixes are not feasible in the short term, update your ToS to allow subscribers to access their own sessions programmatically. This would let users build tools that work while these issues are addressed. Paying $200/mo for a product we can't reliably use, with no workaround permitted, is not acceptable.
This issue has 4 comments on GitHub. Read the full discussion on GitHub ↗