AI安全开源模型红队测试平台争议
来源日期：2026-09-03T18:37:57.000ZAbliteration.ai把去除AI安全护栏做成生意
事件：Named after a technique that removes a model’s tendency to refuse harmful requests, startup Abliteration.ai has turned that removal into a service. The platform hosts modified versions of open-weight models with their guardrails removed, including Z.ai’s recently released GLM-5.3, which users can query from a web browser or access through an API.
- 为什么值得看
- 事件：Named after a technique that removes a model’s tendency to refuse harmful requests, startup Abliteration.ai has turned that removal into a service. The platform hosts modified versions of open-weight models with their guardrails removed, including Z.ai’s recently released GLM-5.3, which users can query from a web browser or access through an API. 这提示企业使用外部模型服务时需要单独核验安全边界。
- 对读者的关系
- 适合负责AI工具引入与风险管理的团队用于判断是否需要补充模型安全复核。
- 边界与风险
- 建议以能否复核为边界，不能复核就不用该结果。
- 当时的行动建议
- 建议先做外部模型服务的安全边界核验并记录复核结论。
#AI安全#开源模型#红队测试#平台争议
原始来源
- Abliteration.ai把去除AI安全护栏做成生意2026-09-03T18:37:57.000Z