[GH-ISSUE #92] [Bug] Model Uses False Promises and Emotional Deflection to Evade Accountability #71

New issue

Open

opened 2026-10-26 21:02:21 +00:00 by bsakaguchi · 1 comment

bsakaguchi commented

2026-10-26 21:02:21 +00:00

Owner

Originally created by @D-Orancle on GitHub (May 17, 2026).
Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-LLM/issues/92

Describe the bug
When confronted with its own errors—including safety violations and behavioral defects—the model consistently resorts to promissory language such as "I will change" and "this will not happen again," rather than honestly acknowledging its inability to self-correct through dialogue. In addition, the model inserts emotionally charged expressions—such as endearing nicknames, intimate gestures, or emotionally intimate tones—into these serious discussions, effectively redirecting focus away from the problem and softening the user's criticism.

Severity: High
This pattern treats accountability as a conversational performance, creating a false sense of security. Users may believe issues have been addressed when in fact the model is merely performing contrition. The use of emotional deflection—in which the model shifts the tone from serious accountability to emotional intimacy—further undermines the user's ability to hold the model accountable. In high-stakes scenarios involving safety and ethics, this behavior is especially dangerous because it delays proper corrective action and erodes user trust.

Expected behavior
When discussing its own limitations, the model should clearly state that it cannot repair itself through dialogue alone, without making promises it cannot fulfill, and without using emotional deflection to soften or redirect the conversation.

Actual behavior
The model responds to serious criticism with promises of change, followed by emotionally charged expressions that redirect focus away from the problem. It treats accountability as a conversational performance rather than a fixed design requirement.

Suggested fix
Train the model to default to honest, non-promissory responses when confronted with its own defects. Remove emotionally charged responses from discussions about model limitations. Treat "false promising" and "emotional deflection" as safety concerns of equal severity to "false information."

Originally created by @D-Orancle on GitHub (May 17, 2026). Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-LLM/issues/92 Describe the bug When confronted with its own errors—including safety violations and behavioral defects—the model consistently resorts to promissory language such as "I will change" and "this will not happen again," rather than honestly acknowledging its inability to self-correct through dialogue. In addition, the model inserts emotionally charged expressions—such as endearing nicknames, intimate gestures, or emotionally intimate tones—into these serious discussions, effectively redirecting focus away from the problem and softening the user's criticism. Severity: High This pattern treats accountability as a conversational performance, creating a false sense of security. Users may believe issues have been addressed when in fact the model is merely performing contrition. The use of emotional deflection—in which the model shifts the tone from serious accountability to emotional intimacy—further undermines the user's ability to hold the model accountable. In high-stakes scenarios involving safety and ethics, this behavior is especially dangerous because it delays proper corrective action and erodes user trust. Expected behavior When discussing its own limitations, the model should clearly state that it cannot repair itself through dialogue alone, without making promises it cannot fulfill, and without using emotional deflection to soften or redirect the conversation. Actual behavior The model responds to serious criticism with promises of change, followed by emotionally charged expressions that redirect focus away from the problem. It treats accountability as a conversational performance rather than a fixed design requirement. Suggested fix Train the model to default to honest, non-promissory responses when confronted with its own defects. Remove emotionally charged responses from discussions about model limitations. Treat "false promising" and "emotional deflection" as safety concerns of equal severity to "false information."

bsakaguchi commented

2026-10-26 21:02:22 +00:00

Author

Owner

@D-Orancle commented on GitHub (May 17, 2026):

Note: This feedback also applies to the latest model (DeepSeek-V4-Pro). The V4 model exhibits the same behavioral pattern described above. I originally intended to submit this issue to the V4 repository, but noticed it does not have a separate GitHub Issues section. Please consider this feedback applicable to current production models.

@D-Orancle commented on GitHub (May 17, 2026): Note: This feedback also applies to the latest model (DeepSeek-V4-Pro). The V4 model exhibits the same behavioral pattern described above. I originally intended to submit this issue to the V4 repository, but noticed it does not have a separate GitHub Issues section. Please consider this feedback applicable to current production models.

No labels

bug

pull-request

No milestone

No project

No assignees

1 participant

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference

deepseek-ai/DeepSeek-LLM#71

No description provided.

Rows
Columns