|
Work item:
|
XSTR.sem-AIA
|
|
Subject/title:
|
Security evaluation methods for Artificial Intelligence agent
|
|
Status:
|
Under study
|
|
Approval process:
|
Agreement
|
|
Type of work item:
|
Technical report
|
|
Version:
|
New
|
|
Equivalent number:
|
-
|
|
Timing:
|
2027-12 (Medium priority)
|
|
Liaison:
|
ITU-T SG13, ITU-T SG20, SG21, ETSI SAI, ISO/IEC, IETF, CEN/CENELEC TC21
|
|
Supporting members:
|
China Telecom, Vivo Mobile Communication, China Mobile, CAICT
|
|
Summary:
|
With the rapid advancement of artificial intelligence (AI) technologies, AI agents are increasingly being utilized across a wide array of fields. By leveraging essential capabilities such as perception, cognition, planning, memory and action, AI agents are able to deliver efficient and intelligent services to users. However, as AI agent capabilities advance and their application expand, security risks have begun to emerge, underscoring the importance of research into the security of AI agents.
While existing standards have already developed a standardized description of security needs for AI agents - by analyzing potential threats and offering corresponding security requirements and guidelines - there is still a lack of standardized guidance on evaluating the security of AI agents. Without security evaluation, developers may lack insight into potential security vulnerabilities within AI agents, leading to delayed and uninformed security enhancement.
Therefore, to provide standardized guidance for assessing AI agent security, enable early detection of security weaknesses, reduce the costs of security enhancement, and reinforce the security assurance for AI agents, it is essential to study security evaluation methods for AI agents.
This draft technical report focuses on providing methods guidance for assessing the security of AI agents. It will define the security evaluation indicators to the core capabilities of AI agents, including perception & cognition, planning, memory and action, and will provide recommended testing methods corresponding to each security indicator.
|
|
Comment:
|
-
|
|
Reference(s):
|
|
|
Historic references:
|
|
Contact(s):
|
|
| ITU-T A.5 justification(s): |
|
|
|
|
First registration in the WP:
2025-12-11 13:21:28
|
|
Last update:
2026-01-14 15:10:58
|
|
