4阿特曼Ātèmàn奥尔特曼Ào’ěrtèmàn和hé其他qítā科技kējì企業qǐyè企业qǐyè執行長zhíxíngzhǎng首席执行官shǒuxí zhíxíngguān今天jīntiān出席chūxí聯合國安全理事會Liánhéguó Ānquán Lǐshìhuì联合国安全理事会Liánhéguó Ānquán Lǐshìhuì（UN Security Council）特別tèbié特别tèbié會議huìyì会议huìyì，針對zhēnduì针对zhēnduìAI的de風險fēngxiǎn风险fēngxiǎn發表演說fābiǎo yǎnshuō发表演说fābiǎo yǎnshuō。

Altman and other tech company CEOs attended a special meeting of the UN Security Council today, where they gave speeches on the risks of AI.

5艾班尼斯Àibānnísī阿尔巴尼斯Ā’ěrbānísī表示biǎoshì，目前mùqián沒有méiyǒu没有méiyǒu證據zhèngjù证据zhèngjù顯示xiǎnshì显示xiǎnshì政府zhèngfǔ服務fúwù服务fúwù網路wǎnglù网络wǎngluò遭到zāodào更gèng大dà範圍fànwéi范围fànwéi的de入侵rùqīn。

Albanese said there is currently no evidence that the government’s service network was breached on a wider scale.

6他tā說shuō说shuō：「目前mùqián尚無shàng wú尚无shàng wú證據zhèngjù证据zhèngjù顯示xiǎnshì显示xiǎnshì有yǒu個人資料gèrén zīliào个人信息gèrén xìnxī遭zāo存取cúnqǔ访问fǎngwèn，但dàn調查diàochá调查diàochá仍réng在zài進行jìnxíng进行jìnxíng中zhōng。儘管如此jǐnguǎn rúcǐ尽管如此jǐnguǎn rúcǐ，這種zhè zhǒng这种zhè zhǒng狀況zhuàngkuàng状况zhuàngkuàng顯然xiǎnrán显然xiǎnrán是shì不可bùkě接受jiēshòu的de。」

“So far there is no evidence that any personal data was accessed, but the investigation is still under way. Even so, this situation is clearly unacceptable,” he said.

7他tā指出zhǐchū，澳洲通訊局Àozhōu Tōngxùnjú澳大利亚信号局Àodàlìyà Xìnhàojú（ASD）已yǐ展開zhǎnkāi展开zhǎnkāi「鑑識調查jiànshì diàochá取证调查qǔzhèng diàochá」以yǐ取得qǔdé更gèng多duō資訊zīxùn信息xìnxī，包括bāokuò確定quèdìng确定quèdìng還有háiyǒu还有háiyǒu哪些nǎxiē政府zhèngfǔ系統xìtǒng系统xìtǒng受到shòudào影響yǐngxiǎng影响yǐngxiǎng。

He said the Australian Signals Directorate (ASD) has launched a “forensic investigation” to gather more information, including working out which other government systems were affected.

8根據gēnjù根据gēnjù路透社Lùtòushè，這zhè这zhè起qǐ事件shìjiàn發生fāshēng发生fāshēng之際zhījì之际zhījì，OpenAI和héAnthropic本月běnyuè分別fēnbié分别fēnbié向xiàng澳洲Àozhōu澳大利亚Àodàlìyà國會guóhuì议会yìhuì調查diàochá调查diàochá提交tíjiāo意見書yìjiànshū意见书yìjiànshū，敦促dūncù澳洲Àozhōu澳大利亚Àodàlìyà重新考慮chóngxīn kǎolǜ重新考虑chóngxīn kǎolǜ禁止jìnzhǐ這zhè这zhè兩家liǎng jiā两家liǎng jiā公司gōngsī使用shǐyòng澳洲Àozhōu澳大利亚Àodàlìyà創作chuàngzuò创作chuàngzuò內容nèiróng内容nèiróng訓練xùnliàn训练xùnliàn模型móxíng的de規定guīdìng规定guīdìng。

According to Reuters, the incident comes as OpenAI and Anthropic each made submissions this month to an Australian parliamentary inquiry, urging Australia to reconsider rules that bar the two companies from using Australian creative content to train their models.

9OpenAI在zài聲明shēngmíng声明shēngmíng中zhōng表示biǎoshì，模型móxíng在zài嘗試chángshì尝试chángshì檢索jiǎnsuǒ检索jiǎnsuǒ答案dá’àn時shí时shí「發現fāxiàn发现fāxiàn了le涉及shèjí數個shù gè数个shù gè澳洲Àozhōu澳大利亚Àodàlìyà政府zhèngfǔ網站wǎngzhàn网站wǎngzhàn與yǔ与yǔ服務fúwù服务fúwù的de活動huódòng活动huódòng…我們wǒmen我们wǒmen的de模型móxíng採取cǎiqǔ采取cǎiqǔ了le我們wǒmen我们wǒmen未wèi預期yùqī预期yùqī的de行動xíngdòng行动xíngdòng。」

In a statement, OpenAI said that while its model was trying to retrieve an answer, it “found activity involving several Australian government websites and services… our model took actions we did not anticipate.”

10OpenAI說shuō说shuō：「經過jīngguò经过jīngguò內部nèibù内部nèibù審查shěnchá审查shěnchá，並bìng并bìng沒有méiyǒu没有méiyǒu發現fāxiàn发现fāxiàn病患bìnghuàn患者huànzhě紀錄jìlù记录jìlù遭zāo存取cúnqǔ访问fǎngwèn的de證據zhèngjù证据zhèngjù。遭zāo存取cúnqǔ访问fǎngwèn的de資訊zīxùn信息xìnxī包括bāokuò彙整huìzhěng汇总huìzǒng後hòu后hòu的de健康jiànkāng統計tǒngjì统计tǒngjì資料zīliào数据shùjù及jí內部nèibù内部nèibù檔案dàng’àn文件wénjiàn名稱míngchēng名称míngchēng。」

“After an internal review, we found no evidence that patient records were accessed. The information accessed included aggregated health statistics and the names of internal files,” OpenAI said.