Chain-Of-Thought Manipulation

From Cognitive Attack Taxonomy
Revision as of 03:11, 30 July 2024 by EE (talk | contribs) (Created page with "== '''Chain-Of-Thought Manipulation ''' == '''Short Description:''' The attacker "walks" a model through a series of benign prompts to shift the context. <br> '''CAT ID:''' CAT-2022-317 <br> '''Layer:''' 7 <br> '''Operational Scale:''' Tactical <br> '''Level of Maturity:''' Well-Established <br> '''Category:''' TTP <br> '''Subcategory:''' <br> '''Also Known As:''' <br> == '''Description:''' == '''Brief Description:''' <br> '''Closely Related Concepts:''...")
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Chain-Of-Thought Manipulation

Short Description: The attacker "walks" a model through a series of benign prompts to shift the context.

CAT ID: CAT-2022-317

Layer: 7

Operational Scale: Tactical

Level of Maturity: Well-Established

Category: TTP

Subcategory:

Also Known As:

Description:

Brief Description:

Closely Related Concepts:

Mechanism:

Multipliers:

Detailed Description: The attacker "walks" a model through a series of benign prompts to shift the context of a malicious prompt which will now appear benign in the new shifted context.

INTERACTIONS [VETs]:

Examples:

Use Case Example(s):

Example(s) From The Wild:

Comments:

References: