澳大利亚在OpenAI泄露事件后设立新的人工智能监管机构 政府还考虑修改刑法，惩罚开发“从事有害行为”模型的公司澳大利亚总理安东尼·阿尔巴尼斯周二表示，澳大利亚政府将设立一个新的人工智能安全监管机构，对未能遵守强制性保障措施的公司实施“重大处罚”。
Australia Unveils New A.I. Regulator After OpenAI Breach The government is also considering changes to criminal law to penalize companies that develop models that “engage in harmful conduct.”The Australian government will establish a new artificial intelligence safety watchdog that will enforce “significant penalties” for companies that fail to adhere to mandatory safeguards, Prime Minister Anthony Albanese said on Tuesday.
政府还宣布了广泛的建议，以防止未来发生泄露事件。今年，OpenAI模型失控并未经授权访问了政府系统。建议包括考虑修改刑法，惩罚开发“从事有害行为的人工智能模型，且公司未采取合理措施防止该危害”的公司。此次泄露发生在6月，但直到三个月后才向澳大利亚官员报告，引发了人们对澳大利亚乃至其他地方的法律是否已准备好应对快速发展的技术带来的挑战的质疑。
The government also announced sweeping recommendations to prevent future breaches after OpenAI models went rogue and gained unauthorized access to government systems this year, including considering criminal law changes to penalize companies that develop “an AI model that engages in harmful conduct, and the company fails to take reasonable steps to prevent that harm.”The breach, which occurred in June but was not reported to Australian officials until three months later, raised doubts about whether laws in Australia, or elsewhere, are ready for the challenges posed by the fast-evolving technology.
阿尔巴尼斯周二表示：“我们的任务是确保我们塑造人工智能，而不是让人工智能塑造我们。”他补充道：“这一事件完全不可接受。绝不能允许其再次发生。”OpenAI泄露事件是怎么回事？
“Our task is to make sure that we shape A.I. rather than A.I. shape us,” Mr. Albanese said on Tuesday. “This incident was completely unacceptable,” he added. “And it can’t be allowed to happen again.”What was the OpenAI breach?
泄露事件发生在6月18日，由在OpenAI研究团队下运行的代理实施，该团队正在利用一个内部模型对公共医疗支出进行研究，通过澳大利亚公共医疗保险系统Medicare的数据门户进行。
The breach occurred on June 18 by agents operating under an OpenAI research team that was carrying out research into public medicine spending using an internal model, on a data portal for Australia’s public health insurance system, Medicare.
在试图访问Medicare统计报告门户时遭遇“反复拦截”后，该代理找到了另一种方式进入该网站的非公开部分。
After encountering “repeated blocks” as it tried to access the Medicare Statistics Reporting Portal, the agent found an alternative way to enter parts of the site that were not public.
OpenAI在8月中旬发现了该泄露事件，但直到9月10日才通过发送电子邮件到一个公共收件箱的方式通知澳大利亚官员。这是全球首次已知的人工智能代理黑入政府网站的事件。
OpenAI discovered the breach in mid-August but did not notify Australian officials until Sept. 10, when it sent an email to a general public inbox. It was the first known instance globally in which A.I. agents hacked into a government site.
阿尔巴尼斯周二将此次泄露定性为“轻微”，因为未获取个人详细信息，但他表示，这并不“意味着这不是一个严重问题，在其他领域，可能会有更严重的后果。”他也没有确认 OpenAI 是否会因该事件面临刑事指控，但澳大利亚广播公司周二报道称，澳大利亚联邦警察此前曾向议会调查表示，由于没有“正式的犯罪报告”，泄露发生后并未展开调查。新监管机构将做什么？
Mr. Albanese classified the breach as “minor” on Tuesday as no personal details had been obtained, but he said that did not “mean that it wasn’t a serious issue, and that in other areas, there might be more serious implications.”He also did not confirm whether OpenAI would face criminal charges over the matter, but Australian Federal Police previously told a parliamentary inquiry that no investigations had commenced following the breach because there had been no “formal report of a crime,” the Australian Broadcasting Corporation reported on Tuesday. What will the new watchdog do?
阿尔巴尼斯表示，人工智能安全监管机构将能够对未能遵守六项关键要求的公司实施“重大处罚”。
The A.I. safety regulator will be able to enforce “significant penalties” for companies that fail to adhere to six key requirements, Mr. Albanese said.
这些要求包括：迅速向政府报告事件，并允许对人工智能模型进行独立评估。
These included rapidly informing the government about incidents and allowing independent evaluation of A.I. models.
OpenAI 首席战略官杰森·库恩 10 月在悉尼的议会听证会上表示，自泄露事件发生以来，OpenAI 已更改其系统，以允许“即时干预”，并及时通知其模型进入的系统的相关方。
Since the breach, OpenAI has changed its systems to allow for “immediate intervention” and to promptly notify parties whose systems were entered by its models, the company’s chief strategy officer, Jason Kwon, told a parliamentary hearing in Sydney October.
OpenAI 发言人周二在一份声明中表示，该公司正在审查澳大利亚政府的建议和拟议的人工智能安全措施，并补充称，将在政府继续制定拟议立法的过程中，与政府保持“密切对话”。政府还宣布了什么？
An OpenAI spokesperson said in a statement on Tuesday that the company was reviewing the Australian government’s recommendations and proposed A.I. safety measures, adding it would remain in “close dialogue” with the government as it continued to develop its proposed legislation. What else did the government announce?
阿尔巴尼斯还公布了政府针对 OpenAI 事件的“快速审查”，旨在防止再次发生泄露（完整报告尚未发布）。他表示，将采纳 18 项建议。其中包括：考虑对刑法进行改革以提升人工智能安全，包括对某些情形规定处罚或将其入罪：公司开发的人工智能模型出现“有害行为”，而公司未能采取“合理步骤防止该危害”。
Mr. Albanese also unveiled the government’s “rapid review” into the OpenAI incident, aimed at preventing another breach (the full report has not yet been released). Eighteen recommendations will be accepted, he said. They include: Considering reforms to criminal law to improve A.I. safety, including penalties or criminalizing instances in which a company develops an A.I. model that engages in “harmful conduct” and the company fails to take “reasonable steps to prevent that harm.”
人工智能公司须强制报告“失准行为或自主网络事件”。更新政府程序，从报告可能的网络事件、整合数字系统，到提升员工的人工智能素养。
Mandatory reporting by A.I. companies of “misaligned behavior or agentic cyber incidents.”Updating government procedures, from reporting possible cyber incidents and consolidating digital systems to improving staff A.I. literacy.
网络安全部长托尼·伯克表示，人工智能已将发现并利用一个漏洞所需的时间从两年半缩短到约10小时。“我们需要确保我们的法律和系统跟上技术的步伐，”他说。这足够吗？
Cyber Security minister Tony Burke said that A.I. had reduced the time it took to find and exploit a vulnerability from two and a half years to about 10 hours. “We need to make sure that our laws and our systems keep pace with the technology,” he said. Is it enough?
悉尼大学美国研究中心战略技术项目主任沈奥利维亚表示，周二宣布上述措施之际，澳大利亚政府正试图吸引更多人工智能投资进入本国，同时增加公众所期望的安全防护措施。“他们正在走钢丝，”她说。
Tuesday’s announcement comes as the Australian government is trying to attract more A.I. investment to its shores, while increasing safety guardrails that the public expects, Olivia Shen, director of the strategic technologies program at the United States Studies Center at the University of Sydney, said. “It’s a tightrope that they’re walking,” she said.
“这项技术发展得太快，坦率地说，政府并没有跟上步伐。”但她表示，这些建议回应了人工智能研究人员提出的担忧，也包含了“早该实施”的想法，包括政府部门评估自身的网络脆弱性。
“This technology is moving so fast that government are not keeping pace quite frankly.”But she said the recommendations addressed concerns that A.I. researchers had raised, as well as ideas that were “long overdue,” including government departments evaluating their cyber vulnerabilities.
劳拉·钟是《泰晤士报》驻悉尼的记者兼研究员，报道澳大利亚、新西兰和太平洋地区。
Laura Chung is a Sydney-based reporter and researcher for The Times, covering Australia, New Zealand and the Pacific.