OpenAI上周解雇的三名AI安全研究员Jasmine Wang、Tomek Korbak和Mikita Balesni发表公开信，否认公司指控的违反政策处理敏感信息的行为 1。OpenAI方面称他们存在"访问和处理敏感公司信息"的违规 1，但研究员们坚称与第三方AI安全组织的合作是其工作的常规组成部分 1。
Wang明确指出，自己因访问高管邮箱而遭解雇，然而该访问权限是OpenAI为招聘工作特意委派的 1。三人还否认参与向媒体The Information泄露OpenAI最新模型架构信息的指控 1。研究员们警告，这起解雇事件正在产生寒蝉效应，阻碍了公司内部对安全问题的讨论，并破坏了与外部评估方合作所需的开放文化 1。
OpenAI曾在内部备忘录中表示"我们不会因为员工提出关切而解雇他们" 1。被解雇的研究员要求公司坚守其公开承诺，在组织内嵌入第三方安全审计员 1。
Three AI safety researchers recently dismissed by OpenAI have publicly contested the company's allegations against them, challenging the rationale behind their terminations and raising concerns about the broader impact on the organization's safety culture.1 The dismissed researchers—Jasmine Wang, Tomek Korbak, and Mikita Balesni—deny violating company policies regarding the handling of sensitive information, which OpenAI cited as grounds for their removal.1 Wang specifically disputed one of the central accusations, explaining that her access to executive emails was authorized by OpenAI as part of her assigned recruitment responsibilities rather than an unauthorized breach.1
The researchers contend that collaboration with external AI safety organizations constituted a routine and essential component of their work, and they deny involvement in leaking OpenAI's latest model architecture details to The Information.1 They argue that the sudden terminations are creating a chilling effect that undermines the open dialogue necessary for conducting effective safety research within the company.1 According to the researchers, the dismissals are obstructing internal discussions on safety concerns and preventing cooperation with external auditors and evaluation partners.1
The terminated employees are calling on OpenAI to reaffirm its public commitment to embedding third-party safety auditors within the organization.1 An OpenAI internal memo stated that "We do not terminate employees for raising concerns," contradicting the researchers' characterization of their dismissals.1
评论
还没有评论，欢迎留下第一条。