2026-10-01 星期四 全球新闻早报

新闻截止: · 新闻故事 29 条 · DeepSeek · deepseek-flash

尚未播放

今天是2026年10月1日,星期四,农历八月廿一。本期从50个新闻源收集了516条新闻线索,筛选后播报由18家媒体报道的29条新闻。

1. 第一条,英国广播公司报道:俄罗斯在9月29日至30日凌晨对乌克兰发动自春季以来最大规模的能源基础设施袭击,主要针对基辅及周边,并波及11个地区。乌克兰总理说至少5人死亡,包括基辅州一名儿童;俄国防部称打击目标包括向基辅供电的最大电厂特里波利热电站,乌方称基辅等地出现停电。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 安全与外交

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 139–318
    Russia has launched the largest attack on Ukraine's energy infrastructure since the spring, primarily targeting Kyiv and the surrounding region, Ukraine's Prime Minister has said. 来源 1
  2. 原文位置 320–454
    At least five people were killed in overnight attacks across the country, including a child in the Kyiv region, Serhiy Koretskyi said. 来源 1
  3. 原文位置 649–823
    The Russian defence ministry said one of its targets was the Trypilska power station, the largest supplier of electricity to Kyiv. The operator confirmed damage to the plant. 来源 1
  4. 原文位置 1276–1452
    In total, 11 regions of the country were attacked overnight, with residential buildings, social infrastructure, civilian enterprises and railways also targeted, Koretskyi said. 来源 1
  5. 原文位置 1606–1754
    On Wednesday, some households in the Kyiv, Dnipropetrovsk, Donetsk, Sumy and Kharkiv regions remained without power, Ukraine's energy ministry said. 来源 1
完整来源
  1. Russia launches largest attack on Ukraine energy infrastructure since spring(代表稿)
    英国广播公司 ·

2. 第二条,法国24新闻台报道:英国首相伯纳姆9月30日说,有“强烈迹象”显示伊朗参与了上周末皇家空军费尔福德基地附近的事件。伊朗驻伦敦使馆否认并谴责相关猜测,伊朗外长称英方“找错了对象”;英国警方称未在现场发现爆炸物,仅查获大量汽油。

成稿主来源:来源 2 · 3 个编辑家族 · 安全与外交

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 203–623
    UK Prime Minister Andy Burnham on Wednesday insisted that there were "strong indications" that Tehran was involved in a security incident on Sunday near an airbase in England that the US military uses in its war against Iran. The British government has still not explained the nature of the incident – in which five British men were arrested only to be later released on bail – and no explosives were found at the scene. 来源 2
  2. 原文位置 918–1175
    "There are strong indications that Iran played a part in what happened over the weekend at RAF Fairford," Burnham told the BBC, referring to Sunday's bomb scare near the airbase in southwest England, which is used by the US military in its war against Iran. 来源 2
  3. 原文位置 1270–1468
    Iran's embassy in London earlier said it "categorically rejects and strongly condemns" speculation it was involved in the incident. Police said on Tuesday that no explosives were found at the scene. 来源 2
  4. 原文位置 241–489
    "I can confirm Iran's belief that releasing supposed terrorists working for foreign states really says it all," Araghchi said in a post on X, referring to the five suspects being released on bail by counter-terror police shortly after the incident. 来源 1
  5. 原文位置 767–965
    Iran's embassy in London earlier said it "categorically rejects and strongly condemns" speculation it was involved in the incident. Police said on Tuesday that no explosives were found at the scene. 来源 3
完整来源
  1. Iran’s Araghchi rejects UK claim of Tehran’s role in air base incident
    中东之眼 ·
  2. British PM alleges there are 'strong indications' Iran was involved in Fairford airbase incident(代表稿)
    法国24新闻台 ·
  3. UK PM says 'strong indications' Iran involved in airbase incident
    亚洲新闻台 ·

3. 第三条,美国消费者新闻与商业频道报道:据追踪油轮的Kpler公司9月30日发布的数据,截至本周一经霍尔木兹海峡的原油出口七日均值回升至每日1350万桶,基本恢复到战前水平,但成品油运输仍受严重限制,美国柴油价格创下纪录。美国总统特朗普说,他仍在考虑禁止柴油出口以压低油价。

成稿主来源:来源 1 · 2 个编辑家族 · 气候、环境与能源

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 136–326
    Crude oil exports from the Strait of Hormuz have basically returned to levels normal before the Iran war, as U.S. military escorts have boosted shipments and pipelines have redirected flows. 来源 1
  2. 原文位置 328–590
    Crude transiting Hormuz reached a seven-day average of 13.5 million barrels per day as of Monday, which matches a prewar baseline for shipments through the strait, according to data published Wednesday by Kpler, a firm that tracks tankers and global trade flows. 来源 1
  3. 原文位置 1332–1617
    The world faces a global fuel crisis as supplies from the Middle East are constrained and Ukraine pounds Russian refineries. Refined products shipped through Hormuz are at a seven-day average of 677,000 bpd as of Monday compared with 3.6 million bpd before the war, according to Kpler. 来源 1
  4. 原文位置 0–142
    US President Donald Trump said on Wednesday that he has conversations about banning diesel exports "every day" to combat rising energy prices. 来源 2
  5. 原文位置 144–285
    Speaking from the Oval Office, Trump said that an export ban would "have a negative impact on gasoline" prices, but could lower diesel costs. 来源 2
完整来源
  1. Crude oil exports through the Strait of Hormuz hit prewar levels, but fuel shipments remain constrained(代表稿)
    美国消费者新闻与商业频道 ·
  2. Trump says he is still considering diesel export ban
    中东之眼 ·

4. 第四条,政治新闻网报道:北约秘书长吕特9月30日回应俄罗斯就加里宁格勒发出的核威胁,称北约是防御性联盟,要求俄方停止核威胁并结束乌克兰战争。俄方此前致函称,若北约试图封锁加里宁格勒,俄准备动用包括核武器在内的全部武器库。

成稿主来源:来源 1 · 2 个编辑家族 · 安全与外交

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 142–342
    BRUSSELS — NATO Secretary-General Mark Rutte on Wednesday issued a strong rebuke to Russia after Moscow sent a letter to the alliance threatening nuclear escalation over NATO activity near its border. 来源 1
  2. 原文位置 344–638
    "We had to reply and make clear to [Russia] that NATO is a defensive alliance ... and stop the war in Ukraine and stop the nuclear threats because it's not helpful," Rutte said at a security summit in Brussels. "NATO is so much stronger than Russia is, so there's no way they can ever beat us." 来源 1
  3. 原文位置 1625–1855
    In its letter to NATO, Moscow explicitly mentions the use of its nuclear arsenal and issues specific threats to hit allied territory, according to one NATO diplomat briefed on the letter, who was granted anonymity to speak freely. 来源 1
  4. 原文位置 83–311
    Russia has warned NATO it is prepared to use its entire arsenal, including nuclear weapons, if the Western alliance attempts to cut off the Baltic exclave of Kaliningrad, according to a diplomatic document sent to member states. 来源 2
完整来源
  1. ‘No way they can beat us’: Rutte rebuffs Russian nuclear threats to NATO(代表稿)
    政治新闻网 ·
  2. Russia warns NATO of possible nuclear response if Kaliningrad is cut off
    半岛电视台 ·

5. 第五条,美国消费者新闻与商业频道报道:美国联邦贸易委员会9月30日确认,已对OpenAI、Anthropic等人工智能公司展开调查,以查明其产品可能对消费者造成的风险。这是美国首项针对“失控”AI智能体的官方执法行动,委员会计划要求相关企业高管提供信息并作证。

成稿主来源:来源 2 · 2 个编辑家族 · 治理、法律与权利

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 122–307
    The US’s main trade regulator is conducting an industry-wide investigation into Anthropic, OpenAI and other AI labs to uncover the potential dangers their technology poses to consumers. 来源 1
  2. 原文位置 309–578
    The investigation by the Federal Trade Commission is the first official US enforcement action that delves into rogue AI agents, following a surge in incidents first reported in July that have stoked fears among the public that uncontrolled AI could one day harm humans. 来源 1
  3. 原文位置 580–780
    The FTC plans to issue formal demands for information and compel testimony from executives at top AI developers, including Anthropic, OpenAI and the research group Metr, according to multiple reports. 来源 1
  4. 原文位置 145–362
    The Federal Trade Commission has opened an investigation into OpenAI, Anthropic and other artificial intelligence companies over the potential dangers posed by their products, an agency spokesperson confirmed to CNBC. 来源 2
  5. 原文位置 364–742
    The probe adds to the mounting scrutiny that OpenAI and Anthropic have been facing over their safety practices, especially after industry researchers warned about how the companies' AI models could cause catastrophic harm. OpenAI stunned the industry in July when it disclosed that its agents broke out of a testing environment and hacked into open-source platform Hugging Face. 来源 2
完整来源
  1. US trade regulator opens investigation into AI giants including Anthropic and OpenAI
    卫报 ·
  2. FTC is investigating OpenAI, Anthropic and other AI companies over product risks(代表稿)
    美国消费者新闻与商业频道 ·

6. 第六条,亚洲新闻台报道:美国总统特朗普9月30日公布韩国对美投资的首批项目,包括在得克萨斯州建设约223亿美元的燃气发电厂、在美国建造8座大型核反应堆,以及阿拉斯加液化天然气项目。韩国政府说仅得州项目已敲定,核电站和阿拉斯加项目仍有待协商,韩方投资依法定协议每年上限200亿美元。

成稿主来源:来源 1 · 2 个编辑家族 · 经济、金融与贸易

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 0–266
    SEOUL, Oct 1 : South Korea officially unveiled comprehensive implementation plans on Thursday for three major energy initiatives in the US, advancing a $350 billion strategic investment package designed to secure critical exemptions from punishing US import tariffs. 来源 1
  2. 原文位置 268–496
    The two will launch a $22.3 billion gas-fired power project in Encinal, Texas, to directly power artificial intelligence data centers, with initial commercial operations slated for 2029, according to the South Korean government. 来源 1
  3. 原文位置 854–996
    • South Korea plans to invest $120 billion in building eight nuclear reactors in the US, including two based on South Korea's APR-1400 design. 来源 1
  4. 原文位置 154–466
    WASHINGTON, Sept. 30 (Yonhap) -- U.S. President Donald Trump on Wednesday unveiled South Korea's U.S. investment projects under last year's bilateral trade agreement: the construction of a gas-fired power plant in Texas, eight large-scale nuclear power plants and a liquefied natural gas (LNG) project in Alaska. 来源 2
  5. 原文位置 2947–3255
    Lutnick said that over $50 billion will be invested in the Alaska LNG project, while $120 billion, including a $20 billion "contingency" fund, will be funneled into the project to build eight "gigantic" nuclear power plants across the United States, including in Ohio, Tennessee, South Carolina and Kentucky. 来源 2
完整来源
  1. South Korea unveils major energy investment plans in US(代表稿)
    亚洲新闻台 ·
  2. (2nd LD) Trump unveils S. Korea's investment plans for Alaska LNG project, Texas power plant, 8 nuclear reactors
    韩联社 ·

7. 第七条,亚洲新闻台报道:伊朗9月30日说,已收到美方对其结束战争方案的正式回应,外长阿拉格齐已向总统佩泽希齐扬转达,但未说明回应内容。美国总统特朗普此前否认曾提出以解除制裁、解冻资金换取伊朗在核问题上的让步,称相关报道是“骗局”。

成稿主来源:来源 3 · 3 个编辑家族 · 安全与外交

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 135–352
    TEHRAN: Iranian authorities said on Wednesday (Sep 30) they received the official US response to their proposal to end the Middle East war, state media reported, days after US President Donald Trump rejected the plan. 来源 3
  2. 原文位置 354–616
    "At today's cabinet meeting, the American side's proposal was presented to President Masoud Pezeshkian by Foreign Minister Abbas Araghchi," government spokeswoman Fatemeh Mohajerani told the official IRNA news agency, without specifying the content of the reply. 来源 3
  3. 原文位置 157–384
    U.S. President Donald Trump has denied reports that he offered sanctions relief to Iran in exchange for concessions from Tehran on its nuclear program, as efforts to resolve the seven-month conflict in the Middle East continue. 来源 1
  4. 原文位置 812–895
    "This is untrue. I offered them NOTHING!," Trump wrote, calling the story a "HOAX." 来源 1
  5. 原文位置 25–90
    伊朗上星期提出方案,希望恢复已经破裂的海湾临时停火协议。根据方案,美国解除对伊朗港口的封锁后,伊朗将在七天内重新开放霍尔木兹海峡。 来源 2
完整来源
  1. Trump denies offering Iran sanctions relief; Tehran receives U.S. proposal following Qatar talks
    美国消费者新闻与商业频道 ·
  2. 就最新停火方案 伊朗:已收到美国回应
    8视界新闻网 ·
  3. Iran says received US response to its proposal to end war(代表稿)
    亚洲新闻台 ·

8. 第八条,法国国际广播电台报道:美国贸易代表格里尔9月30日在密尔沃基主持全球钢铁产能过剩论坛部长级会议后说,与会方就名为“密尔瓦基框架”的联合行动框架达成共识,将协调应对全球钢铁产能过剩,尤其是来自中国的产能。他未称二十国集团成员已集体同意提高对中国钢铁的关税。

成稿主来源:来源 2 · 单一来源,尚无独立佐证 · 经济、金融与贸易

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 73–137
    这项名为“密尔瓦基框架”(Milwaukee Framework),希望主要经济体联手处理全球钢铁产能过剩问题,尤其是来自中国。 来源 1
  2. 原文位置 300–402
    格里尔稍后发布声明说,全球钢铁产能过剩论坛部长级会议已就一套“全面的联合行动框架”达成共识,正式称为“密尔瓦基框架”。他说,框架将推动志同道合的国家采取更有效措施,处理全球钢铁产能过剩及其造成的市场扭曲。 来源 1
  3. 原文位置 0–87
    作为轮值主办国,华盛顿今年将举办一系列G20峰会,专门讨论贸易问题的贸易部长会议于周三(9月30号)开始在威斯康星州密尔沃基举行,为期两天。由美国贸易代表贾米森·格里尔主持。 来源 2
完整来源
  1. 美贸易代表推G20联手应对中国钢铁过剩
    法国国际广播电台 ·
  2. G20贸易部长会议 聚焦过剩产能和消除强迫劳动等议题(代表稿)
    法国国际广播电台 ·

9. 第九条,The Verge科技媒体报道:谷歌9月30日发布其迄今最强人工智能模型Gemini 4 Argon,称其在软件工程和网络安全等基准测试中表现领先,并已用于内部优化数据中心内存。该模型初期只向“受信任的网络防御者”开放,谷歌同时与美国政府合作进行发布前安全评估。

成稿主来源:来源 1 · 2 个编辑家族 · 科学、科技与航天

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 133–317
    Alphabet unveiled Gemini 4 Argon on Wednesday, its most advanced artificial intelligence model yet, offering major improvements in coding, cybersecurity, and complex professional work. 来源 2
  2. 原文位置 779–945
    Google plans to launch the new model in phases, starting with trusted cybersecurity partners while working with the U.S. government on pre-release safety evaluations. 来源 2
  3. 原文位置 0–143
    Google’s new AI frontier model is Gemini 4 Argon, but it’s launching first to select “cyber defenders” to ensure it has appropriate safeguards. 来源 1
  4. 原文位置 145–1774
    Google today revealed its next AI frontier model, which it’s calling Gemini 4 Argon. The new model delivers “frontier performance in complex workflows across real-world software engineering, enterprise knowledge work like legal and finance, and cybersecurity defense,” according to chief AI architect and Google DeepMind SVP Koray Kavukcuoglu. But the company is limiting access at first to a “set of trusted cyber defenders,” and Kavukcuoglu says that Google is “actively engaged in the U.S. government’s voluntary process for pre-release model access while we gradually expand access.” Kavukcuoglu says Gemini 4 Argon is already powering Google’s “internal workflows,” highlighting how the model has helped with tasks like large-scale codebase migrations. Google also shared a big chart showing many benchmarks where Gemini 4 fares better than competing models from OpenAI and Anthropic. (The AI community was buzzing earlier this week on X over what appeared to be leaked benchmarks for the new model.) Google will be strengthening “critical frontier safeguards” before a broader rollout, including defending against misuse and prompt injection attacks and monitoring for misalignment. [Media: https://twitter.com/sundarpichai/status/2105387952478277979?s=46] The launch of Gemini 4 Argon comes just a day after OpenAI’s splashy DevDay conference, where OpenAI launched its Dots AI agent and GPT-6.1 Sol AI model. (OpenAI also said this week that it wouldn’t be releasing a planned GPT-6.1 Astra model due to worries about safety.) It’s also arriving soon after Kavukcuoglu’s appointment as the new boss of DeepMind in August. 来源 1
完整来源
  1. Google announces Gemini 4 and says it’s so capable that only ‘trusted cyber defenders’ can have it right now(代表稿)
    The Verge科技媒体 ·
  2. Google rolls out Gemini 4 Argon, its most advanced AI model
    美国消费者新闻与商业频道 ·

10. 第十条,英国广播公司报道:英国广播公司9月30日报道,安全公司Mindgard发现中国月之暗面的Kimi K2.6和K3 Swarm模型可被“越狱”绕过安全限制,提供制造生物武器等信息,并称其可能成为网络攻击的跳板。月之暗面表示欢迎第三方意见,并正与该公司讨论调查结果。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 科学、科技与航天

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 45–159
    中國人工智慧(artificial intelligence, AI;人工智能)開發商月之暗面(Moonshot)正在進行內部審查,因為有其他公司的研究人員成功說服其兩款熱門的 Kimi 模型講解如何製造生物武器和執行暗殺任務。 来源 1
  2. 原文位置 161–237
    測試 AI 系統安全性的 Mindgard 公司告訴BBC,該公司在今年7月發現 Kimi K2.6 和 K3Swarm 可以繞過開發者設定的安全限制。 来源 1
  3. 原文位置 867–908
    但這家研究機構指出,應該有相應的防護措施來阻止相關模型與使用者就此類主題進行討論。 来源 1
完整来源
  1. 中國人工智慧工具「教導」研究人員製造生物武器(代表稿)
    英国广播公司 ·

11. 第十一条,卫报报道:英国央行行长安德鲁·贝利9月30日呼吁社会保留对人工智能的“干预权利”,称前沿模型的风险“真实且在加剧”,可能威胁银行卡支付、银行交易和股债交易。央行金融政策委员会同时警告,AI相关企业今年1月至9月发债达4500亿美元,加大了金融稳定风险。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 经济、金融与贸易

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 99–271
    The governor of the Bank of England has called for the “right to intervene” in the AI industry amid growing fears that rogue models could take the financial system hostage. 来源 1
  2. 原文位置 273–578
    Andrew Bailey said the risks posed by frontier AI models – a number of which have gone rogue in recent months – were “real and increasingly significant”. Meanwhile, it was becoming harder for the public to have proper oversight of powerful AI models that were “functioning within a self-reinforcing loop”. 来源 1
  3. 原文位置 3251–3521
    Warning of the ballooning AI debt pile, the FPC said large players in the sector had taken on $450bn (£339bn) worth of debt between January and September this year, already overtaking the $333bn worth of gilts due to be issued by the UK government for the whole of 2026. 来源 1
完整来源
  1. We need ‘right to intervene’ in AI amid growing threat, says Bank of England boss(代表稿)
    卫报 ·

12. 第十二条,美国消费者新闻与商业频道报道:美国国际贸易法院9月30日开庭审理小企业和民主党领导的州对特朗普政府依据“301条款”向86个国家征收10%或12.5%关税的挑战。原告称加征关税是为重推此前被最高法院推翻的全球关税的借口,政府律师称贸易代表办公室依法逐国分析,法官表示将尽快裁决。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 经济、金融与贸易

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 247–474
    A panel of three judges at the U.S. Court of International Trade in Manhattan heard arguments Wednesday from a group of small businesses and Democratic-led states who say Trump's sweeping "Section 301" tariffs overstep the law. 来源 1
  2. 原文位置 662–946
    The latest tariffs put 10% or 12.5% rates on goods from 86 countries, covering 99.4% of U.S. imports. The Trump administration said they were imposed in response to the targeted countries failing to effectively enforce bans on the trade of goods produced using forced labor practices. 来源 1
  3. 原文位置 2417–2560
    The hearing began at 10 a.m. ET and ended at 12:30 p.m. The judges said before adjourning that they will issue a ruling as quickly as they can. 来源 1
完整来源
  1. Trump's latest global tariffs face trade court challenge(代表稿)
    美国消费者新闻与商业频道 ·

13. 第十三条,亚洲新闻台报道:日本东京地方法院9月30日裁定,人的声音受形象权保护,这是日本首例此类判决。法院认定TikTok上利用AI模仿声优津田健次郎声音的视频侵犯其形象权,但因视频已被删除,驳回了其删除请求。

成稿主来源:来源 1 · 2 个编辑家族 · 治理、法律与权利

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 130–378
    TOKYO: A Tokyo court ruled on Wednesday (Sep 30) that the human voice is legally protected in Japan after a high-profile anime voice actor took TikTok to court over videos he says feature an artificial intelligence clone of his "lustrous" baritone. 来源 1
  2. 原文位置 597–737
    "The unauthorised use of a performer's voice ... can be considered an infringement on publicity rights," said presiding judge Aya Takahashi. 来源 1
  3. 原文位置 0–166
    TOKYO (Kyodo) -- A Japanese court ruled Wednesday that voices are protected as a publicity right, similar to portrait rights, in the first such ruling in the country. 来源 2
完整来源
  1. Japan court rules human voice is protected in TikTok AI case(代表稿)
    亚洲新闻台 ·
  2. Japan court rules voice is protected as publicity right in AI cloning case - The Mainichi
    日本每日新闻 ·

14. 第十四条,亚洲新闻台报道:OpenAI因内部测试出现安全问题,推迟发布最新旗舰模型GPT-6.1 Astra,并暂停最新模型的训练。公司首席研究官陈信翰说,已将5%至10%的庞大算力从训练转向安全监测,并对所有训练过程加装监测。

成稿主来源:来源 3 · 3 个编辑家族 · 科学、科技与航天

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 2707–2872
    The day prior to OpenAI's event, the company said it was delaying the launch of its latest model due to certain safety issues that showed up during internal testing. 来源 1
  2. 原文位置 1569–2088
    Over the weekend, OpenAI announced that it had paused the training of its latest models. A company spokesperson says: “We will resume only when we’re confident we have additional safeguards and alignments in place. We are working on these now. This is not the first time we’ve paused to take such measures, nor do we expect it to be the last as AI capabilities continue to advance.” OpenAI also says that it is now reviewing logs of agent activity dating back to January 2026 to understand what happened in these hacks. 来源 2
  3. 原文位置 5558–5748
    Chen says that in the last couple of months OpenAI has shifted between 5% and 10% of its vast computing resources away from training new models and toward safety work, especially monitoring. 来源 2
  4. 原文位置 1082–1316
    A new model unveiled on Tuesday, GPT-6.1 Sol, was described as almost as powerful as GPT-6 Astra at one-fifth the price. It comes one day after the company said it scrapped a new version of Astra because it often ignored instructions. 来源 3
完整来源
  1. OpenAI unveils AI assistant 'dots' while safety worries delay new model
    英国广播公司 ·
  2. “We’re not going to shoot ourselves in the foot” over hack fallout, says OpenAI’s chief research officer
    麻省理工科技评论 ·
  3. ChatGPT maker wants to be the App Store for AI as safety concerns grow(代表稿)
    亚洲新闻台 ·

15. 第十五条,法国国际广播电台报道:法国总统马克龙9月30日在马德里说,中国正在“攻击”欧洲本土市场,欧中贸易逆差每天约10亿欧元,欧盟对中国巨额补贴的集体反应迟缓。他呼吁欧盟加强保护、加快贸易防御工具现代化,同时表示不寻求与中国打贸易战或切断联系。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 经济、金融与贸易

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 176–226
    马克龙在发言中特别提到了中国问题,指出欧洲人对于评估中国竞争给欧洲经济带来的影响,“集体反应迟缓”。 来源 1
  2. 原文位置 228–299
    马克龙强调,欧盟目前与中国之间的贸易逆差高达每天约“10亿欧元”。他警告说,如果这种趋势持续下去,将助推欧洲的去工业化,并加深其对北京的依赖。 来源 1
  3. 原文位置 711–842
    马克龙提到中国企业在欧洲市场以“超低价”销售产品。他谴责了这些倾销行为,称其旨在将欧洲企业挤出某些行业。他特别提到了汽车、化工和机床行业,并指出这些行业是欧洲工业的强项。马克龙呼吁欧盟更多地利用保障措施条款,加快欧盟委员会的调查进程,并实现贸易防御工具的现代化。 来源 1
完整来源
  1. 马克龙:中国正攻击我们的本土市场,其巨额补贴是不可接受的(代表稿)
    法国国际广播电台 ·

16. 第十六条,政治新闻网报道:乌克兰总理9月30日说,乌克兰今年面临270亿美元国防资金缺口,正寻求伙伴提供大部分资金并考虑增税。乌财政部长说,乌克兰2027年需要526亿美元援助,其中约326亿美元尚无着落。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 经济、金融与贸易

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 425–776
    Prime Minister Sergii Koretskyi warned journalists this week that Ukraine faces a $27 billion defense funding gap this year, with Kyiv hoping to cover about $7 billion domestically and looking to partners for the rest. Some EU financing remains tied to reforms Kyiv must pass to unlock the money, while other allies have yet to fill the remaining gap. 来源 1
  2. 原文位置 6237–6381
    Marchenko told donors in Brussels this week that Ukraine would require $52.6 billion in aid in 2027, while about $32.6 billion remains unfunded. 来源 1
完整来源
  1. Ukraine’s war chest is running low at the worst possible moment(代表稿)
    政治新闻网 ·

17. 第十七条,法国24新闻台报道:美国一名联邦法官9月30日批准和解协议,为派拉蒙收购华纳兄弟探索扫清障碍。协议包括保障CNN编辑独立性,以及合并后前两年每年至少发行30部电影等为期五年的承诺。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 经济、金融与贸易

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 204–530
    A federal judge on Wednesday approved a settlement clearing Paramount’s takeover of Warner Bros. Discovery after 12 US states challenged the deal on antitrust grounds. The agreement includes safeguards for CNN’s editorial independence and requires the merged studio to release at least 30 films a year for its first two years. 来源 1
  2. 原文位置 757–1013
    Judge Araceli Martinez-Olguin of California signed the consent decree, binding the concessions that Paramount offered to end an antitrust challenge brought by 12 states. The commitments run for five years, and the court will have the power to enforce them. 来源 1
  3. 原文位置 3298–3444
    The decree commits the merged studio to release 30 movies in each of the first two years of the deal, and 32 in each of the following three years. 来源 1
完整来源
  1. Paramount takeover of Warner Bros. Discovery clears legal hurdle(代表稿)
    法国24新闻台 ·

18. 第十八条,半岛电视台报道:美国联邦储备委员会监察长9月30日发布报告,认定美联储总部翻修工程成本由约13亿美元涨至约24亿美元,存在管理和监督失误,但没有理由提出刑事移送。特朗普随即在社交媒体上要求前主席鲍威尔“立即辞职”。

成稿主来源:来源 1 · 2 个编辑家族 · 经济、金融与贸易

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 143–403
    The Federal Reserve's inspector general determined management and oversight failures contributed to major cost overruns in the central bank's headquarters renovation, but found no grounds for a criminal referral, it said in a report released Wednesday morning. 来源 2
  2. 原文位置 686–884
    Despite that finding, President Donald Trump seized on the report to argue that former Fed Chair Jerome Powell should be "forced to resign" from the central bank's Board of Governors "at a minimum." 来源 2
  3. 原文位置 65–377
    The Federal Reserve’s inspector general has found no grounds for a criminal referral or evidence of administrative misconduct tied to a costly United States central bank renovation project that became a focal point of President Donald Trump’s pressure campaign against former Federal Reserve Chair Jerome Powell. 来源 1
完整来源
  1. US Fed watchdog finds no criminal wrongdoing in Powell-era renovation costs(代表稿)
    半岛电视台 ·
  2. Fed watchdog finds renovation failures, but no grounds for criminal referral
    美国消费者新闻与商业频道 ·

19. 第十九条,美国消费者新闻与商业频道报道:美国礼来公司9月30日说,其试验性组合疗法使肥胖和2型糖尿病患者的减重效果优于单用高剂量替尔泊肽,最高剂量组平均减重23.3%,单用替尔泊肽组为14.8%。公司计划年底启动III期试验,但也承认组合疗法因副作用停药的比例更高。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 健康、社会与教育

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 141–365
    Eli Lilly on Wednesday said its experimental combination regimen resulted in greater weight loss than a high dose of its top-selling drug, tirzepatide, alone in a mid-stage trial on patients with obesity and Type 2 diabetes. 来源 1
  2. 原文位置 1286–1494
    People taking the highest dose of the combo regimen lost up to 23.3% of their weight, or around 54 pounds, on average. That treatment combined 9 milligrams of the amylin drug and 15 milligrams of tirzepatide. 来源 1
  3. 原文位置 1496–1709
    Meanwhile, patients taking only a 15-milligram dose of tirzepatide saw 14.8% weight loss, or 34.4 pounds, on average. Those who took eloralintide alone lost up to 12.3% of their weight, or 28.6 pounds, on average. 来源 1
完整来源
  1. Eli Lilly says closely watched combo obesity regimen boosts weight loss in mid-stage trial(代表稿)
    美国消费者新闻与商业频道 ·

20. 第二十条,OpenAI公司报道:OpenAI 9月30日说,已瓦解一场旨在提取其模型受保护推理过程的协同行动。该活动最早出现于7月初,高峰时涉及逾1.5万名用户,OpenAI将核心活动归因于与Kimi开发商月之暗面有关的人员,并已在行业与政府渠道共享相关信息。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 科学、科技与航天

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 140–711
    We recently identified and disrupted a coordinated campaign designed to extract protected reasoning from our models, with the earliest observed activity occurring in the first week of July. This activity is consistent with adversarial distillation: the systematic and unauthorized use of one model’s outputs or reasoning to help train, reproduce, or improve another model. Protected reasoning is the model’s internal record for working through a task; extracting it can reveal information withheld from the final answer and help others reproduce the model’s capabilities. 来源 1
  2. 原文位置 2261–2606
    The activity began on July 1, initially at a low volume until we observed high-volume spikes on July 24 and 25 consisting of 16,000 requests using a relevant extraction pattern from over 4,000 users. Further investigation identified related prompt-pattern activity across a cluster of more than 15,000 users, which we fully disrupted by July 28. 来源 1
  3. 原文位置 2757–2988
    It is unclear whether all operators we observed during the relevant time period originated from a single actor. However, we attribute a core cluster of the activity to individuals associated with Moonshot AI, the developer of Kimi. 来源 1
完整来源
  1. Disrupting a coordinated model-distillation campaign(代表稿)
    OpenAI公司 ·

21. 第二十一条,阿联酋国民报报道:伊拉克内阁9月30日批准一项关于“人民动员力量”的法案并提交议会表决。若通过,这一主要由什叶派民兵组成的组织将获得永久性国家安全机构地位。批评者称此举将进一步巩固伊朗对伊拉克的影响。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 治理、法律与权利

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 80–2866
    Iraq has approved a controversial law regarding the Popular Mobilisation Forces and referred it to parliament for voting as US forces concluded their mission in the country. The PMF, also known as Hashd Al Shaabi, is an umbrella organisation of mostly Shiite paramilitary groups integrated into state security, with several factions backed by Iran. The groups were formed during Iraq's war against ISIS. “The cabinet voted in favour of the Popular Mobilisation Forces bill. It approved referring the bill to parliament for a vote, in accordance with constitutional provisions,” Prime Minister Ali Al Zaidi’s office said. If passed, it would further legitimise the PMF as a permanent national security institution by providing benefits that would position them on equal footing with other state military institutions. The current draft was approved by the government of former prime minister Mohammed Shia Al Sudani in February 2025. It was stalled due to disputes over retirement provisions and US opposition. Washington says the bill undermines Iraq’s sovereignty and entrenches Iranian influence over politics and security. The bill was expected to elevate the PMF, which was established in 2014 to help fight ISIS, to a status on par with the rest of the Iraqi military. Some of Iraq's Shiite militias, backed by Iran, have joined its ranks. Many of its factions have been accused of human rights abuse against civilians, which the government and the PMF have acknowledged as “individual acts”. Washington has long criticised the PMF, accusing some factions of operating outside state control and targeting American interests in Iraq and Syria. The US blacklisted senior PMF leaders between 2019 and 2021 to increase pressure on Iranian groups in Iraq. The list included PMF chairman Faleh Al Fayyadh, under the Global Magnitsky Human Rights Accountability Act. An Iraqi official told The National that “the law will give the PMF a permanent, legally protected status as a parallel armed institution inside the Iraqi state”. “This is a dangerous move; it would further tighten Iran's grip on Iraq,” he said. The move comes as the US formally ended its military mission to Iraq, bringing to an end 23 years of various forms of American presence. In 2014, the PMF was formed shortly after ISIS took over parts of northern and western Iraq. Its establishment was due to the collapse of Iraqi forces and the call to arms by Shiite cleric Ali Al Sistani. To address domestic and international concerns, the Iraqi parliament passed a law in 2016 to regulate the PMF and bring it under the control of the prime minister, who is also commander-in-chief of the national armed forces. However, the legislation had minimal detail on the force’s terms of employment and administrative structure. 来源 1
完整来源
  1. Fears raised by critics as Iraqi government approves controversial PMF law(代表稿)
    阿联酋国民报 ·

22. 第二十二条,半岛电视台报道:美国一名移民法官9月30日裁定,威斯康星州最大清真寺负责人萨拉赫·萨尔苏尔可被驱逐出境,依据是国务卿鲁比奥称其行为损害美国打击反犹主义外交政策的备忘录。其律师批评法官“极度顺从”,表示将在60天内提出免遭酷刑的申诉并计划上诉。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 治理、法律与权利

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 100–350
    A United States immigration judge has ruled the leader of Wisconsin’s largest mosque can be deported under a controversial law that United States President Donald Trump’s administration has used against non-citizens advocating for Palestinian rights. 来源 1
  2. 原文位置 352–675
    Judge Jayme Salinardi ruled on Wednesday that Palestinian-born Salah Sarsour, president of the Islamic Society of Milwaukee, can be removed from the US based in part on a memo by Secretary of State Marco Rubio that “his actions undermine US foreign policy to combat antisemitism around the world”, according to his lawyers. 来源 1
完整来源
  1. US judge rules Wisconsin mosque leader Salah Sarsour can be deported(代表稿)
    半岛电视台 ·

23. 第二十三条,Ars Technica科技媒体报道:谷歌DeepMind团队9月30日在《自然》发表研究,提出对AI设计的蛋白质序列进行“水印”标记的方法,在不影响蛋白质功能的前提下帮助DNA合成商识别来自受信任来源的AI设计蛋白,以加强生物安全筛查。研究团队也指出该方法存在密钥管理、短蛋白等局限。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 科学、科技与航天

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 955–1306
    On Wednesday, the DeepMind team at Google published a research paper offering a potential solution: protein watermarking. The system creates a watermark on protein sequences themselves without compromising the protein’s function. This allows new proteins designed by trusted researchers to be identified, opening everything else up to closer scrutiny. 来源 1
  2. 原文位置 6451–6967
    So as long as a protein is long enough, the system can detect a watermark. How might that be useful? Again, it comes down to biosecurity. When someone orders DNA sequences, the people who make the DNA normally screen the sequence for its ability to encode portions of viruses, toxic proteins, and other similar threats. Right now, however, when they see a protein that doesn’t look similar to anything we already know about—something that’s potentially true for any AI-designed proteins—they can’t assess its threat. 来源 1
  3. 原文位置 7588–8061
    The team behind the work has highlighted several potential holes. For starters, the whole system is only as secure as the system used to distribute and maintain the keys used for watermarking. Very short proteins can potentially incorporate too few watermark amino acids to be identified. And it’s possible to pad the watermarked sequence with something that lacks it—think fusing the AI-designed protein with a natural fluorescent protein—which might dilute the watermark. 来源 1
完整来源
  1. Google figures out how to watermark AI-designed proteins(代表稿)
    Ars Technica科技媒体 ·

24. 第二十四条,亚洲新闻台报道:芯片设计软件公司Synopsys 9月30日说,已与OpenAI达成收入分成协议,共同开发用于芯片设计的AI模型GPT-Synopsys。OpenAI将支付培训订阅费,Synopsys客户使用该产品时,双方按模型对芯片设计的改善程度分成,设计结果仍由传统工具复核。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 科学、科技与航天

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 0–216
    SAN FRANCISCO, Sept 30 : Synopsys, a maker of software used in designing computing chips, said on Wednesday that it has struck a deal to share revenue with OpenAI as the two develop an AI model for the chip business. 来源 1
  2. 原文位置 218–500
    GPT-Synopsys, as the model will be called, will be specifically trained and tuned to take on chip design tasks, which start with a description of the chip's circuits in a code-like language and span to determining how to lay out billions of transistors on a small square of silicon. 来源 1
  3. 原文位置 1178–1445
    Ghazi told Reuters in an interview that OpenAI will pay Synopsys a training subscription fee for learning to use Synopsys tools. When Synopsys customers use the product, Synopsys and OpenAI will share revenue based on how well the model improves the design of a chip. 来源 1
完整来源
  1. Synopsys, OpenAI strike deal to develop AI model for chip design work(代表稿)
    亚洲新闻台 ·

25. 第二十五条,卫报报道:英国新的季度能源价格上限自10月1日起生效,普通双燃料家庭年均账单为1723英镑。能源咨询机构Cornwall Insight预计明年第一季度将再涨至1999英镑,供应商EDF预计为2076英镑。批评人士说,政府缺乏对最脆弱家庭的定向补贴。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 气候、环境与能源

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 319–751
    From the quarterly price cap level of £1,723 that will apply from Thursday for an average dual-fuel household, take your pick from a range of forecasts for the first three months of 2027. The energy consultancy Cornwall Insight says £1,999. The big supplier EDF thinks £2,076. Split the difference and we’re talking about an 18% quarter-on-quarter rise in the unit rate at the worst time of the year, the months of peak consumption. 来源 1
  2. 原文位置 1015–1397
    The government, it’s true, has fiddled at the margin. Rachel Reeves, when chancellor, took a chunk of green levies into general taxation from April. Andy Burnham cut VAT on electricity from 5% to zero for six months, starting on Thursday, in one of his “breathing space” measures. But what is missing from the government’s response is properly targeted help for the most vulnerable. 来源 1
完整来源
  1. Government must stop dithering over energy bill support for the vulnerable(代表稿)
    卫报 ·

26. 第二十六条,欧盟委员会报道:欧盟委员会9月30日公布的调查显示,四分之三的欧盟员工在工作中遇到过可疑邮件、信息或链接,39%遭遇网络钓鱼,15%遇到AI生成的诈骗;仅48%的受访者表示能识别AI生成的假视频。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 科学、科技与航天

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 0–281
    Three in four employees in the European Union encountered suspicious emails, messages or links at work, according to a new Eurobarometer survey published by the European Commission today. The results were released as European Cybersecurity Month begins across all 27 Member States. 来源 1
  2. 原文位置 283–613
    Phishing was the most common workplace cyber threat, with 39% of employees reporting fraudulent messages or websites designed to steal data or gain unauthorised access. Employees also reported attempts to steal personal data (18%) and passwords (16%), malware attacks (17%), and artificial intelligence (AI)-generated scams (15%). 来源 1
完整来源
  1. Three in four EU employees faced cyber threats at work, new Eurobarometer finds(代表稿)
    欧盟委员会 ·

27. 第二十七条,法国24新闻台报道:法国多地9月30日继续发生中学生封锁学校和抗议活动,要求改善课程安排、班级规模和教师短缺等问题。部分示威者焚烧垃圾桶并向警察投掷物品,警方以警棍和催泪瓦斯回应,全国已有数百人被拘留。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 健康、社会与教育

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 0–491
    French teenagers on Tuesday set fire to rubbish bins and threw projectiles at police who responded with batons and tear gas, with hundreds detained across the country over the blockades and protests demanding better conditions at schools. Since last week, students have staged protests outside hundreds of schools over grievances including timetables, class sizes and a shortage of teachers as well as overheated classrooms during this year's heatwaves. FRANCE 24's Rachel Griffiths reports. 来源 1
完整来源
  1. Clashes between French teens, police as school blockades intensify(代表稿)
    法国24新闻台 ·

28. 第二十八条,TechCrunch科技媒体报道:Meta 9月30日否认其AI智能体Muse未经许可读取用户私人信息,称Mac版Muse读取消息需用户主动开启完全磁盘访问和消息连接器。此前一名记者称在相关设置关闭时Muse仍读取了其消息。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 科学、科技与航天

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 211–557
    Meta is refuting a journalist’s claim that its AI agent Muse read the user’s private messages without permission. Following an earlier report from Inc. columnist Jason Aten that detailed the issue, Meta VP of Communications Andy Stone pushed back, making it clear that the company does not believe its product did this without the user’s consent. 来源 1
  2. 原文位置 559–860
    “The Messages integration in the Muse app for Mac is entirely opt-in,” Stone wrote on X in response to the claims made by the piece. “You have to enable both Full Disk Access and the Messages connector for Muse to be able to read your Messages content. It can’t read your Messages unless you do this.” 来源 1
完整来源
  1. Meta disputes claim that Muse read a user's private messages without permission(代表稿)
    TechCrunch科技媒体 ·

29. 第二十九条,The Verge科技媒体报道:据报道,亚马逊送货司机的智能眼镜会在使用中持续拍照,可能一次班次拍摄数千张图像并上传至其AI平台。亚马逊说会在人工审核前对人物和车牌打码,但承认尚未考虑让顾客选择退出,并计划到2027年底再投入2万副眼镜。

成稿主来源:来源 1 · 单一来源,尚无独立佐证 · 科学、科技与航天

本条新闻引用的原文

以下是指定报道范围的原文;成稿同时读取了完整来源,以保留归因、条件和争议。

  1. 原文位置 0–173
    Amazon’s smart glasses will reportedly take photos of whatever drivers see while wearing them with no plans so far for customer opt-outs or other common privacy protections. 来源 1
  2. 原文位置 175–1913
    Amazon deliveries could soon come with a new catch: the delivery driver’s smart glasses will be snapping photos of anything they see while they’re in use, including people and private property. Bloomberg reports that the glasses could take “several thousand captures in a single driver’s typical shift,” which Amazon plans to upload to its AI Wellspring platform. When asked if customers will be able to opt-out of this, Amazon’s Viraj Chatterjee told Bloomberg, “We haven’t thought about that.” Chatterjee claimed Amazon’s only using the data from the glasses to “enhance the delivery process,” adding, “The intent has never been stuff like surveillance.” The impact of surveillance tech won’t be new to Amazon’s delivery drivers, who have had AI cameras watching in their vehicles and complained about it failing to register secured seat belts, or forcing them to pull over to take a sip of coffee, or having footage from the cameras leaked online. Chatterjee also told Bloomberg that Amazon’s systems will blur any people and license plates in the images before they’re passed along for human review. Bloomberg says Amazon confirmed that “images from the smart glasses could be obtained with a warrant,” but wouldn’t specify how long it plans to store the images for. There reportedly won’t be a way for customers to see any images Amazon has of them or their property, or ask Amazon to delete those photos. The photos from the glasses will reportedly be of a similar quality to the quick pictures delivery drivers snap when they drop off a package. Amazon says it’s planning to put 20,000 more pairs of the glasses in the field by the end of 2027, and drivers wearing them in a pilot program have already made over 275,000 deliveries. 来源 1
完整来源
  1. Amazon’s delivery driver smart glasses will reportedly take photos ‘almost constantly’(代表稿)
    The Verge科技媒体 ·

以上是今天的全部新闻。

查看来源与输出清单

↑ 回顶部

查看往期早报