您的位置：首页 → 单机游戏 → 角色扮演 → 怪物猎人世界新装备介绍怪物猎人世界新装备新系统一览

英雄

US cyberattack push risks fueling 'privateer-style' cyber piracy, disrupts global cyberspace order, expert warns_我的网站

无证之罪

A |

Photo: VCG
The US government is encouraging American companies to launch their own cyberattacks against criminal hackers, a move that certain US media claimed would bring the US closer to the practices of some of its chief cyber adversaries, including China and Russia. A Chinese expert slammed the claim as factually unfounded, saying that the US move would seriously disrupt the global cyberspace order. Washington has long projected its own conduct onto other countries, inventing adversaries to justify the militarization of cyberspace, the weaponization of artificial intelligence (AI) and even a privateer-style model of cyber piracy, said the expert.
Under a national security memorandum that US President Donald Trump signed late Wednesday, select companies would work with the Justice and Homeland Security Departments to strike foreign cybercriminal groups with hacks under certain conditions. The attacks would allow both surveillance of the criminal networks and specific types of hacking operations that could lead to disruption, manipulation or destruction of information systems and networks, including virtual and physical infrastructure, according to a report from the New York Times.
A CNN report called the move "a major policy shift that experts say could come with legal risk for the companies." It also noted that "companies can use cyber tools to surveil foreign criminals and disrupt their networks under the 'control and oversight' of the new federal program."
The US move is an attempt, under the guise of combating cybercrime, to accelerate its plan to issue "cyber privateering licenses" to private entities. Its core objective is to exploit its advantages in the cybersecurity and intelligence sectors and rapidly turn AI-enhanced cyberattack capabilities into tools for securing strategic interests and undermining the security of other nations, Xiao Xinguang, chief software architect from Chinese cybersecurity company Antiy, told the Global Times on Friday.
Analysts noted that some US media outlets routinely drag China into the frame and slander it whenever Washington unveils relevant policies, all to manufacture pretexts to legitimize America's own policies.
The New York Times report claimed that "to some extent, the new order would align the US more with some of its chief cyberantagonists, including China and Russia, where spy agencies have long relied on contract hackers working in the private sector to further their national security missions."
Xiao noted that the US has long projected its own practices onto other countries, manufacturing imaginary adversaries as scapegoats and falsely attributing to China the operational models of its own intelligence agencies, contractor networks and revolving-door system. By applying double standards and leveling unfounded accusations, Washington creates pretexts for militarizing cyberspace, weaponizing AI and even developing a cyber-privateering model.
In reality, certain US corporations have long supported US intelligence agencies in intelligence collection and offensive cyber operations, and this has been an established practice for years, Xiao said.
"Now, the US is bringing these previously covert practices into the open. As the trend develops, more US cybersecurity firms and intelligence contractors could operate under so-called privateering authorizations, posing as non-state actors while carrying out intrusions, data theft, sabotage and other disruptive cyberattacks against countries around the world," said Xiao.
。

B |

IT之家 8 月 23 日消息，据南华早报报道，一家获得 OpenAI 投资的美国人工智能初创公司，基于中国 AI 实验室月之暗面（Moonshot AI）的 Kimi K3 构建了自己的首个内部模型。这凸显出，随着 AI 模型开发成本不断飙升，越来越多西方科技公司开始转向采用中国的开放权重模型。

C | 总部位于旧金山的法律科技公司 Harvey 当地时间周四宣布，其新模型 Harvey Tenet 是在开放权重模型 Kimi K3 的基础上进行后训练得到的。Harvey 的知名投资方还包括红杉资本和 Andreessen Horowitz。该公司表示，这套系统在复杂法律工作中实现了“业界领先”的性能。这对 Harvey 而言意味着一次重大转变。该公司为大型国际律师事务所和企业客户提供服务，此前主要专注于针对法律应用，对 Anthropic、OpenAI 和谷歌等美国 AI 巨头的闭源专有模型进行定制。AI 政策研究员 Simon Hedlin 周五在社交媒体上表示，Harvey 的转变是开放权重模型价值的一个“绝佳例子”。开发者可以利用这类模型，基于特定行业或企业的数据进行后训练，从而提高准确性并降低推理成本。所谓后训练，是指利用专业数据集对通用基础模型进行进一步优化，使其能够更擅长完成特定任务。Hedlin 表示：“我们目前仍处于探索高能力开放权重模型究竟能够实现什么的非常早期阶段。”他还称：“遗憾的是，美国在前沿开放权重模型的开发方面已经落后了。”Harvey 成立于 2022 年夏季，并在今年 3 月的一轮融资中获得了 110 亿美元（IT之家注：现汇率约合 741.74 亿元人民币）的估值。

D | Harvey 在周四发布的博客文章中表示，公司过去六个月的研究重点，是利用开放权重模型打造“前沿法律智能”，并让律师事务所能够构建和部署专用模型。

E | 根据 Harvey 的说法，经过全面法律数据集训练后，Harvey Tenet 在一系列复杂、长周期的法律智能体任务中，表现优于其底层基础模型以及美国的前沿 AI 系统，其中包括 Fable 5 和 GPT-5.6 Sol。

F | Harvey 表示，这种方法还提高了成本效率。该公司指出，开放权重模型本身通常具有更低的每 Token 成本，而 Harvey 还通过减少推理过程中消耗的 Token 数量，进一步降低了成本。

G | Harvey 称，Harvey Tenet 的训练历时两个月，使用了约 150 块英伟达 B300 GPU。与此同时，据《The Information》周四援引美国电信巨头 AT&T 副总裁 Mark Austin 的报道称，包括英伟达 Nemotron 系列在内的开源模型，目前已经处理了 AT&T 员工发起的约 40% 的 AI 查询。

。

H |

Current article:http://www.bianhajieleigoukusunmen.pics/1aah/qusugf.doc

Published on:11:02:56

文章评论

查看所有0条评论>>