开启新的安全漏洞奖励计划,以识别AI滥用风险. OpenAI launches new Safety Bug Bounty program to identify AI abuse risks.

开源AI已经推出了一个新的安全漏洞奖励计划,以识别人工智能滥用风险,补充其现有的安全程序. OpenAI has launched a new Safety Bug Bounty program to identify AI abuse risks, complementing its existing security program. 该倡议主要关注诸如劫持人工智能代理人以造成伤害,泄露专有信息以及绕过反自动化控制等问题. The initiative focuses on issues like hijacking AI agents for harm, exposing proprietary information, and bypassing anti-automation controls. 研究人员可以获得可行的发现奖励,但一般内容政策的绕过被排除在外,除非它们造成直接伤害. Researchers can earn rewards for actionable findings, though general content-policy bypasses are excluded unless they cause direct harm.