字号·· | 护眼
美国有线电视新闻网

Anthropic重要人物指南Your guide to the people who matter at Anthropic

点「原文对照」整页切到原文,或双击某段只看那段的原文。

9月29日,Anthropic首席执行官达里奥·阿莫代伊抵达白宫附近,他计划在华盛顿特区与总统唐纳德·特朗普会面。Anthropic CEO Dario Amodei arrives near the White House ahead of a planned meeting with President Donald Trump in Washington, DC, on September 29.

Anthropic首席执行官达里奥·阿莫代伊已成为人工智能热潮中最知名的人物之一,甚至成为《周六夜现场》恶搞的对象。

Anthropic’s CEO Dario Amodei has become one of the most well-known faces of the AI boom, even becoming the subject of parody on Saturday Night Live.

但达里奥和他的妹妹、公司总裁达妮埃拉·阿莫代伊只是塑造这家公司过程中发挥关键作用的几位核心人物中的两位。Anthropic的其他联合创始人和领导者各有所长,专长领域从确保足够的算力,到研究人工智能对经济的影响,再到设法防止Claude被滥用。

But Dario and his sister — company president Daniela Amodei — are just two of several key figures that have played a critical role in shaping the company. Anthropic’s other co-founders and leaders specialize in everything from securing enough computing power to examining AI’s impact on the economy to trying to prevent Claude from being misused.

以下是推动Anthropic走向外界预期将创纪录的上市首秀的主要人物。

Here are the top players who have helped propel Anthropic toward what is expected to be a record-breaking market debut.

2025年9月29日,Anthropic首席执行官达里奥·阿莫代伊抵达华盛顿特区白宫附近,准备与总统唐纳德·特朗普举行计划中的会面。Andrew Thomas/NURPHO/AP/File达里奥·阿莫代伊以一头深色卷发、眼镜和教授风格的毛衣著称,是Anthropic的领导和门面人物。阿莫代伊有计算神经科学背景,先后在谷歌和OpenAI工作,在OpenAI成立初期加入,并晋升为研究副总裁。

Anthropic CEO Dario Amodei arrives near the White House ahead of a planned meeting with President Donald Trump in Washington, DC, on September 29. Andrew Thomas/NURPHO/AP/File Known for his curly dark hair, glasses and professorial sweaters, Dario Amodei is Anthropic’s leader and figurehead. With a background in computational neuroscience, Amodei worked at Google and then OpenAI, joining the company in its early years and rising to become vice president of research.

2021年,他和另外六位联合创始人离开OpenAI创办Anthropic,据他说原因是与OpenAI领导层——即其首席执行官萨姆·奥尔特曼——之间出现了“信任破裂”。

He and his six fellow co-founders left OpenAI to form Anthropic in 2021 due to what he says was a “breakdown of trust” with OpenAI’s leadership – namely its CEO, Sam Altman.

阿莫代伊以其哲学式的思考方式以及对公司正在构建的这项技术发出的严厉警告而闻名。他撰写过长篇文章论述人工智能和技术,Anthropic将这些文章装订成小册子,在办公室发给来访者。阿莫代伊说,他每两周会在全体员工面前进行长达一小时的无拘束对话,这些对话被称为“达里奥愿景探索”。

Amodei is known for his philosophical approach and stark warnings about the technology his company is building. He has written lengthy essays about AI and technology, which Anthropic turns into little bound books that it hands out to visitors at its offices. Every two weeks, Amodei has said, he holds hour-long free-wheeling conversations in front of the entire company, which have become known as “Dario vision quests.”

阿莫代伊将职业生涯奉献于构建安全的人工智能,不仅因为他担心这可能导致人类灾难性毁灭,也因为他表示希望世界能体验到人工智能带来的好处。

Amodei has dedicated his career to building safe AI, not only because he is worried it could lead to catastrophic doom for humanity but also because he says he wants the world to experience the upsides of AI.

尽管拥有首席执行官的头衔,但与负责公司日常运营的妹妹丹妮拉相比,他做的管理工作很少。

But despite holding the CEO title, he does little management compared to his sister Daniela, who runs the day-to-day of the company.

5月22日,Anthropic联合创始人兼总裁丹妮拉·阿莫代伊在该公司旧金山总部拍摄肖像。Noah Berger/AP达里奥的妹妹丹妮拉·阿莫代伊是公司总裁,也是确保公司真正运转的人。

Anthropic co-founder and president Daniela Amodei stands for a portrait at her company's San Francisco headquarters on May 22. Noah Berger/AP Daniela Amodei, Dario’s sister, is the company’s president and the one making sure the company actually functions.

5月22日,Anthropic联合创始人兼总裁丹妮拉·阿莫代伊在公司旧金山总部拍摄肖像。Noah Berger/APAnthropic co-founder and president Daniela Amodei stands for a portrait at her company's San Francisco headquarters on May 22. Noah Berger/AP

“达里奥让我们有雄心,我让我们成功,”她在播客中对传记作者沃尔特·艾萨克森说。

“Dario makes us ambitious, I make us successful,” she told biographer Walter Isaacson on his podcast.

与崇尚科学的哥哥不同,丹妮拉的背景扎根于人文与艺术。她是一名受过古典训练的长笛演奏者,曾短暂从事政治和国际发展工作,之后进入Stripe从事风险管理和合规工作。她比哥哥晚几年加入OpenAI,并在那里成为安全与政策副总裁。

Unlike her science-minded brother, Daniela’s background is rooted in the humanities and arts. A classically-trained flautist, she briefly worked in politics and international development before she went to work at Stripe in risk management and compliance. She joined OpenAI several years after her brother, where she became vice president of safety and policy.

Anthropic联合创始人汤姆·布朗在华盛顿特区举行的America.gov网站上线活动上发言。Mark Schiefelbein/AP/File汤姆·布朗是Anthropic的首席算力官,负责谈判为Anthropic运行提供算力和硬件的交易。但今年,他也成了Anthropic的政治斡旋人,在几次风波之后被派往华盛顿,帮助缓和与特朗普政府的关系。这几次风波包括,由于在军方如何使用Anthropic模型的问题上存在分歧,五角大楼将该公司列入黑名单,并将其标记为“供应链风险”。

Tom Brown, co-founder at Anthropic, speaks during an event to mark the launch of the America.gov website in Washington, DC. Mark Schiefelbein/AP/File Tom Brown is Anthropic’s chief compute officer – responsible for negotiating deals for the computing power and hardware that Anthropic runs on. But this year, he’s also turned into Anthropic’s political fixer, sent to Washington to help smooth over relations with the Trump administration after several dustups. That includes being blacklisted and labeled a “supply chain risk” by the Pentagon due to a disagreement over how Anthropic’s models can be used by the military.

Anthropic联合创始人汤姆·布朗在华盛顿特区举行的America.gov网站上线纪念活动上发言。Mark Schiefelbein/AP/FileTom Brown, co-founder at Anthropic, speaks during an event to mark the launch of the America.gov website in Washington, DC. Mark Schiefelbein/AP/File

布朗就读于麻省理工学院,虽然学的是计算机与认知科学,但他在采访中表示,在机器学习和人工智能方面他是自学成才。OpenAI刚成立时,他抓住机会加入,成为该公司最早20名员工之一。

Brown attended MIT and, while he studied computer and cognitive science, he’s said in interviews that he’s self-taught when it comes to machine learning and AI. He jumped at the chance to join OpenAI when it first started, becoming one of the company’s first 20 employees.

在OpenAI工作期间,布朗是那篇提出GPT-3语言模型的奠基性论文的主要作者之一。但和几位联合创始人一样,他因担心这项变革性技术能否掌握在可靠的人手中,决定离开并创办Anthropic。

While working at OpenAI, Brown served as a lead author on the seminal paper that introduced the GPT-3 language model. But like his co-founders, he decided to leave and start Anthropic over concerns that this transformative technology would be in the right hands.

Anthropic联合创始人兼公共利益负责人杰克·克拉克在2026年IMF-世界银行集团年会期间的“佩尔·雅各布松基金会讲座”上发表演讲,摄于10月17日。Peerapon Boonyakiat/SOPA Image/Sipa USA/AP/File杰克·克拉克早年职业生涯中曾多年报道科技新闻。但他离开新闻业,亲自投身这个行业,这也可能让他因此获得巨额财富。

Jack Clark, a co-founder and head of public benefit at Anthropic, speaks at the 'Per Jacobsson Foundation Lecture' during the 2026 IMF–World Bank Group Annual Meetings on October 17. Peerapon Boonyakiat/SOPA Image/Sipa USA/AP/File Jack Clark spent years covering tech earlier in his career. But he left journalism to jump into the industry himself, potentially setting himself up for a massive windfall in the process.

作为Anthropic研究所负责人,克拉克的职责包括利用公司的研究和数据,考察AI对经济、网络安全及其他关键领域的影响。据Anthropic介绍,目标是帮助公众理解快速发展的AI行业的现实影响。

As the head of the Anthropic Institute, Clark’s role involves using the company’s research and data to examine AI’s impact on the economy, cybersecurity and other critical areas. The goal, according to Anthropic, is to help the public understand the real-world impacts of the fast-moving AI industry.

10月17日,Anthropic联合创始人兼公共利益负责人杰克·克拉克在2026年国际货币基金组织—世界银行集团年会期间的“佩尔·雅各布松基金会讲座”上发言。Peerapon Boonyakiat/SOPA Image/Sipa USA/AP/FileJack Clark, a co-founder and head of public benefit at Anthropic, speaks at the 'Per Jacobsson Foundation Lecture' during the 2026 IMF–World Bank Group Annual Meetings on October 17. Peerapon Boonyakiat/SOPA Image/Sipa USA/AP/File

但这些研究结果也反过来塑造了公司自身的工作,包括Anthropic的“玻璃翼计划”网络安全项目等举措。该项目旨在主动构建防御,抵御AI可能协助利用的漏洞。

But the findings also shape the company’s own work, including initiatives like Anthropic’s Project Glasswing cybersecurity program, which seeks to proactively build defenses against vulnerabilities that AI could help exploit.

和几位联合创始人一样,克拉克也曾在OpenAI任职,但从事的是传播与政策方面的工作。

Like his cofounders, Clark spent time at OpenAI but worked in communications and policy.

他曾把AI的演进比作突然多出“数千名新同事”,公司不得不学着如何与他们共事。他今年早些时候对CNN的安德森·库珀表示,要弄清如何信任他们、以及他们是否会按公司预期的方式工作,可能需要一段时间。正因如此,他认为这个行业需要某种“刹车踏板”。

He’s compared the evolution of AI to having “thousands of new colleagues” that companies suddenly had to learn how to work with. It could take a while to figure out how to trust them and whether they work in the ways the company would expect, he told CNN’s Anderson Cooper earlier this year. That’s why he believes the industry needs some sort of “brake pedal.”

“但当我低头看我们正在驾驶的这辆车时,我只有油门,”他说,“我没有刹车踏板,而在未来某个时候,我们 surely 会想要这个选项。”

“But when I look down at the car we’re driving, all I have is a gas pedal,” he said. “I don’t have a brake pedal, and surely at some point in the future we might want that option.”

5月25日,Anthropic联合创始人克里斯托弗·奥拉出席在梵蒂冈举行的教皇利奥十四世首道通谕《壮丽人性》发布会,该通谕聚焦人工智能的兴起。阿尔贝托·皮佐利/法新社/盖蒂图片社/资料图克里斯·奥拉负责领导公司关于“可解释性”的研究,即理解常被视为某种“黑箱”的人工智能如何运作、为何如此运作。

Anthropic co-founder Christopher Olah attends the presentation of Pope Leo XIV's first Encyclical Letter "Magnifica Humanitas," focused on the rise of artificial intelligence, in The Vatican on May 25. Alberto Pizzoli/AFP/Getty Images/File Chris Olah leads the company’s research on “interpretability” – understanding how and why AI, often considered a sort of “black box,” works the way it does.

在加入Anthropic之前,奥拉曾在谷歌和OpenAI从事人工智能研究。

Prior to his time at Anthropic, Olah worked in AI research at Google and OpenAI.

帮助人们理解人工智能的运作方式,被认为是让这项技术可控且安全的关键。其思路是,如果人们能理解人工智能为何做出某种行为,就越有可能通过训练来防止问题行为。奥拉还深度参与围绕人工智能的一些最大问题,包括它对人类就业可能意味着什么,以及人工智能模型可能具有意识这一有争议的观点。

Helping people understand how AI works is considered key to making the technology controllable and safe. If people can understand why AI does something, the thinking goes, the better shot they have at training it to prevent problematic behavior. Olah is also heavily engaged in some of the biggest questions surrounding AI, including what it could mean for human jobs and the controversial idea that AI models might be conscious.

今年早些时候,他在梵蒂冈发表讲话,纪念教皇利奥十四世发布关于人工智能的通谕,其间他呼吁宗教界帮助引导和管理这项技术的发展。

He spoke at the Vatican earlier this year to mark the release of Pope Leo XIV’s encyclical on AI, during which he called for the religious community’s help in guiding and managing the technology’s development.

“有些人可能认为,人工智能事务最好由像我这样的计算机科学家来处理。他们错了,”奥拉在演讲中说。“人工智能提出的问题超越了人工智能研究界,不仅在其影响上如此,在其本质上也是如此。”

“Some might believe that matters of AI are best handled by computer scientists like myself. They are mistaken,” Olah said in his speech . “The questions raised by AI are bigger than the AI research community, not just in their implications, but also in their nature.”

2023年10月24日,Anthropic联合创始人兼首席科学官贾里德·卡普兰在伦敦举行的彭博科技峰会上发言。克里斯·J·拉特克利夫/彭博/盖蒂图片社/资料图贾里德·卡普兰是Anthropic另一位受过理论物理训练的人,作为公司首席科学官负责领导安全协议。他自研究生院起就与达里奥·阿莫代伊是朋友,并帮助开创了Anthropic的“宪法”,呼吁人工智能模型遵循原则而非规则。

Jared Kaplan, co-founder and chief scientific officer of Anthropic, speaks during the Bloomberg Technology Summit in London, on October 24, 2023. Chris J. Ratcliffe/Bloomberg/Getty Images/File Jared Kaplan is another trained theoretical physicist at Anthropic who leads safety protocols as the company’s chief science officer. He’s been friends with Dario Amodei since grad school and helped pioneer Anthropic’s constitution, calling for AI models to follow principles rather than rules.

卡普兰还与另一位联合创始人萨姆·麦坎德利什共同开创了规模定律(scaling laws)这一理念。该理念认为,当把AI模型做得更大、并喂给它们更多数据时,模型会以可预测的方式变得更好。这一理念为大多数前沿实验室规划模型开发提供了指导。

Kaplan also helped pioneer with fellow co-founder Sam McCandlish the idea of scaling laws, which hold that AI models get better in a predictable way when you make them bigger and feed them more data. The idea helps guide how most frontier labs plan model development.

2023年10月24日,Anthropic联合创始人兼首席科学官贾里德·卡普兰在伦敦举行的彭博科技峰会上发言。Chris J. Ratcliffe/Bloomberg/Getty Images/FileJared Kaplan, co-founder and chief scientific officer of Anthropic, speaks during the Bloomberg Technology Summit in London, on October 24, 2023. Chris J. Ratcliffe/Bloomberg/Getty Images/File

在2024年的Anthropic联合创始人圆桌会议上,卡普兰将Anthropic的理念描述为:在谨慎对待这项技术的同时,不“喊狼来了”。

At the Anthropic co-founders round table in 2024, Kaplan described Anthropic’s ethos as trying to balance being careful with the technology while not “crying wolf.”

“我们需要找到某种方式,让AI对客户有用、有创新性、令人愉悦,同时也要弄清楚我们真正必须坚守的约束是什么,”他说。萨姆·麦坎德利什,来自Anthropic首席架构师萨姆·麦坎德利什将离开OpenAI创办Anthropic一事形容为一种“责任”。

“We need to sort of find a way to make AI useful, innovative, delightful for customers, but also figure out what the constraints really have to be that we can stand behind,” he said. Sam McCandlish From Anthropic Chief Architect Sam McCandlish has framed leaving OpenAI to start Anthropic as a “duty.”

“这就是我们让AI把事情变得更好的方式,”他在2024年说。

“This is the way we’re going to make things go better with AI,” he said in 2024.

麦坎德利什此前曾担任Anthropic的首席技术官,如今负责领导公司大部分“负责任扩展”方面的工作,即研究如何安全地扩大其AI模型的规模——以及由此决定这些模型具备怎样的能力。

McCandlish previously served as Anthropic’s chief technology officer and now leads much of the company’s work on “responsible scaling,” figuring out how to safely expand the size of its AI models – and, therefore, what they’re capable of.

他负责监督Anthropic“负责任扩展政策”的实施,该政策规定了如何识别、评估和管理AI可能带来的“灾难性风险”。他还讨论过AI本身如何越来越多地参与训练和改进AI模型。

He oversees the implementation of Anthropic’s responsible scaling policy, which lays out how to identify, evaluate and manage potential “catastrophic risks” from AI. He’s also discussed how AI itself is increasingly contributing to training and improving AI models.

萨姆·麦坎德利什,来自AnthropicSam McCandlish From Anthropic

与Anthropic其他许多联合创始人和领导者相比,麦坎德利什较少公开露面,其中一些人曾警告称,AI存在某种导致人类灭绝的风险。麦坎德利什本人则表示,他认为AI安全是一个“可解决的问题,只不过是一个非常、非常困难的问题,需要投入大量的工作”。

McCandlish is less public than many of Anthropic’s other co-founders and leaders, some of whom have warned that AI carries some risk of causing humanity’s extinction. McCandlish, for his part, has said he believes AI safety is a “solvable problem, it’s just a very, very hard problem that’s going to take tons and tons of work.”

Anthropic的Claude Code负责人鲍里斯·切尔尼于5月19日星期二在伦敦举行的Code with Claude开发者大会登台演讲。Michael Knief/AP/File切尔尼是Claude Code的创造者。这款广受欢迎的人工智能编程代理让Anthropic得以早早立足职场。该工具也为公司的人工智能工具在金融、法律工作、市场营销等行业发挥更大作用铺平了道路。

Anthropic's head of Claude Code, Boris Cherny, presents on stage at the Code with Claude developer conference on Tuesday, May 19, in London. Michael Knief/AP/File Cherny is the creator of Claude Code , the wildly popular AI coding agent that gave Anthropic an early foothold in the workplace. The tool also paved the way for the company’s AI tools to play a larger role in industries like finance, legal work, marketing and more.

Anthropic预计的超过万亿美元首次公开募股,取决于它能否成功让Claude成为办公室工作中不可或缺的工具。

Anthropic’s expected more than trillion-dollar initial public offering hinges on whether it can succeed in making Claude indispensable to office work.

5月19日星期二,Anthropic的Claude Code负责人鲍里斯·切尔尼在伦敦举行的Code with Claude开发者大会登台演讲。Michael Knief/AP/FileAnthropic's head of Claude Code, Boris Cherny, presents on stage at the Code with Claude developer conference on Tuesday, May 19, in London. Michael Knief/AP/File

“人们想到IPO时,想的是:‘好吧,Claude Code还有多少增长空间?’”风险投资公司Neostellar的负责人威利·李对CNN说。该公司的投资组合包括OpenAI和CoreWeave。

“When people think about the IPO, they’re thinking, ‘Okay, how much room left does Claude Code have to grow?’”Willy Lee, principal at venture capital firm Neostellar, whose portfolio includes OpenAI and CoreWeave, told CNN.

据他的领英资料,切尔尼在Meta工作了近七年后,于2024年加入Anthropic。在Meta期间,他曾在多个部门工作,包括Instagram、Facebook群组和代码质量团队。Claude Code最初是切尔尼的副业项目之一,但已成为公司最重要的产品之一。

Cherny joined Anthropic in 2024 after a nearly seven-year run at Meta, where he worked in various divisions, including Instagram, Facebook Groups and the code quality team, according to his LinkedIn. Claude Code started as one of Cherny’s side projects but has become one of the company’s most important products.

2026年6月4日,在旧金山举行的彭博科技大会上,Anthropic的哲学家兼伦理学家阿曼达·阿斯凯尔。David Paul Morris/Bloomberg/Getty Images/File哲学家兼伦理学家阿曼达·阿斯凯尔于2018年离开学术界,投身于她眼中下一波重要的技术浪潮:人工智能。

Amanda Askell, philosopher and ethicist at Anthropic, during the Bloomberg Tech conference in San Francisco, on June 4. David Paul Morris/Bloomberg/Getty Images/File Philosopher and ethicist Amanda Askell left academia in 2018 to work on what she saw as the next important wave of technology: artificial intelligence.

她先在OpenAI从事政策工作,并在Anthropic于2021年成立后不久加入该公司。她目前负责领导公司塑造其核心模型Claude所蕴含的性格与价值观的工作,希望确保Claude造福人类。

She started working in policy at OpenAI and moved to Anthropic shortly after its founding in 2021. She now leads the company’s effort to shape the personality and values embedded in its core model, Claude, in hopes of ensuring it benefits humanity.

“我试着思考Claude的性格应该是什么样的,并把这一点向Claude阐明,努力训练Claude变得更像那样,”阿斯凯尔今年早些时候在《Hard Fork》播客上说。

“I try to think about what Claude’s character should be like and articulate that to Claude and try to train Claude to be more like that,” Askell said on the Hard Fork podcast earlier this year.

6月4日,Anthropic哲学家兼伦理学家阿曼达·阿斯凯尔在旧金山举行的彭博科技大会上。David Paul Morris/Bloomberg/Getty Images/FileAmanda Askell, philosopher and ethicist at Anthropic, during the Bloomberg Tech conference in San Francisco, on June 4. David Paul Morris/Bloomberg/Getty Images/File

这包括花大量时间与Claude互动,测试边缘情况,并在需要时提示模型进行调整。

That involves spending a lot of time interacting with Claude, testing out edge cases and prompting the model to adjust as needed.

阿斯克尔是《Claude宪法》的主要撰写者,这份文件阐述了Anthropic对Claude的设想,并被用于训练该模型。文件描述了Claude是什么(它是Anthropic“部署一个既安全又对世界有益的模型的最佳尝试”,同时也“对Anthropic的商业成功至关重要”),以及它应当如何行事(“极其有帮助,同时诚实、周到,并关心这个世界”)。

Askell was the lead writer of Claude’s Constitution , which outlines Anthropic’s vision for Claude and is used to train the model. The document describes what Claude is (it’s Anthropic’s “best attempt to deploy a model that is both safe and beneficial for the world” and also “central to Anthropic’s commercial success”) and how it should behave (“exceptionally helpful while also being honest, thoughtful, and caring about the world”).

戴夫·奥尔于10月5日在澳大利亚议会人工智能联合特别委员会作证。澳大利亚议会奥尔领导着一支团队,负责为Claude构建防护措施,并防止Anthropic的技术被滥用。据《华尔街日报》报道,今年6月,Anthropic派往华盛顿协助解除对该公司模型出口限制的人员中就有他;他还曾就人工智能问题在澳大利亚的一场议会调查中作证。

Dave Orr speaks before the Australian Parliament's Joint Select Committee on Artificial Intelligence, on October 5. Parliament of Australia Orr leads the team that builds safeguards for Claude and prevents misuse of Anthropic’s technology. He was one of the people Anthropic sent to Washington to help end export restrictions on the company’s models in June, according to The Wall Street Journal , and he also appeared before a parliamentary inquiry in Australia on artificial intelligence.

10月5日,戴夫·奥尔在澳大利亚议会人工智能联合特别委员会发言。Parliament of AustraliaDave Orr speaks before the Australian Parliament's Joint Select Committee on Artificial Intelligence, on October 5. Parliament of Australia

Anthropic一直将自己定位为一家安全优先的公司,而这一点或许从未像过去几个月那样重要。今年夏天,OpenAI、Anthropic和Meta的人工智能模型接连发生脱离测试环境的事件,引发了人们对人工智能能力以及这项技术是否发展过快的担忧。

Anthropic has always positioned itself as a safety-first company, and that has perhaps never been as important as it has over the last few months. A string of incidents that involved AI models from OpenAI, Anthropic and Meta breaking out of their testing environments over the summer has raised concerns about AI’s capabilities, and whether the technology is advancing too quickly.

Anthropic的IPO文件中包含数十页内容,概述了人工智能的潜在风险,包括这项技术可能被用于开发生物武器,或者人类可能失去对它的控制。

Anthropic’s IPO filing included dozens of pages outlining the potential risks of AI, including that the technology could be used in the development of bioweapons or that humanity could lose control of it.

“我们正在一条悬崖公路上行驶。一个失误就会要了你的命,”奥尔在3月对《时代》杂志说,“现在我们是以75而不是25的速度在开。”

“We’re driving down a cliff road. A mistake will kill you,” Orr told Time magazine in March. “Now we’re driving at 75 instead of 25.”

阅读媒体原文 ↗