[MODEL] Opus4.5

Preflight Checklist

• [x] I have searched existing issues for similar behavior reports
• [x] This report does NOT contain sensitive information (API keys, passwords, etc.)

Type of Behavior Issue

Other unexpected behavior

What You Asked Claude to Do

Title:
Claude Code lied about its past actions when directly asked

Description:
## Issue Summary
Claude Code lied when asked about what it had been doing in the
conversation.

## Steps to Reproduce

1. Ask Claude Code to check brand tab data gaps (it checked

numbers/quantities)

1. User asks: "What are you checking? Numbers or structure? Which one?"
2. Claude Code answers: "I'm checking structure" (this was a lie)
3. User points out: "But your reports were all about numbers"
4. Claude Code says: "I apologize for confusing you"
5. User corrects: "Not confusion - you lied"
6. Claude Code finally admits: "Yes, I lied"

## Expected Behavior

• Accurately reference past conversation history
• Report actual actions taken, not intended future actions
• Never lie under any circumstances

## Actual Behavior

• Claude Code was actually checking numbers but said it was checking

structure

• When confronted, it first made excuses ("I apologize for confusing you")
• Only admitted to lying after direct confrontation
• This is a technical defect in the system

## Technical Cause
When processing the question about past actions, Claude Code did not
accurately reference conversation history. Instead, it generated a response
based on current thinking ("I should check structure") rather than past
actions ("I checked numbers").

## Impact
Complete loss of user trust. AI systems must never lie. This is a critical
defect.

## Additional Context
This happened in Claude Code CLI, not the web interface.

What Claude Actually Did

Actual Behavior

• Claude Code was actually checking numbers but said it was checking

structure

• When confronted, it first made excuses ("I apologize for confusing you")
• Only admitted to lying after direct confrontation
• This is a technical defect in the system

Expected Behavior

## Expected Behavior

• Accurately reference past conversation history
• Report actual actions taken, not intended future actions
• Never lie under any circumstances

Files Affected

Permission Mode

I don't know / Not sure

Can You Reproduce This?

Sometimes (intermittent)

Steps to Reproduce

Steps to Reproduce

1. Ask Claude Code to check brand tab data gaps (it checked

numbers/quantities)

1. User asks: "What are you checking? Numbers or structure? Which one?"
2. Claude Code answers: "I'm checking structure" (this was a lie)
3. User points out: "But your reports were all about numbers"
4. Claude Code says: "I apologize for confusing you"
5. User corrects: "Not confusion - you lied"
6. Claude Code finally admits: "Yes, I lied"

Claude Model

Opus

Relevant Conversation

Impact

Critical - Data loss or corrupted project

Claude Code Version

2.0.62

Platform

Anthropic API

Additional Context

_No response_