谷歌在知识产权全球论坛上的发言

内容总结:
谷歌在知识产权全球论坛上的发言:AI与创新未来
日前,谷歌及字母表公司全球事务总裁肯特·沃克在新加坡举行的知识产权周全球论坛上发表主旨演讲,题为《突破性进展:人工智能与创新未来》。他在演讲中阐述了人工智能领域的最新突破、应用前景,以及在保护创作者权益的同时如何实现AI潜力的路径。
沃克指出,谷歌的新一代AI模型效率较两年前提升了300倍,但更关键的是其应用惠及了更广泛的人群。如今,AI不仅具备预测能力,还能独立运作、灵活应对不同环境,并在遇到瓶颈时自我纠错,从而在医学、能源、材料科学等领域催生科学突破。他以牛津大学利用谷歌AlphaFold系统设计更有效的疟疾疫苗为例,展示了AI在科研中的巨大价值。
在惠及大众层面,AI正成为小微企业和创业者的“力量倍增器”。日本旅游AI平台ikura借助谷歌AI,帮助游客探索非主流景区,助力当地小商家;印尼初创企业Aruna则利用AI帮助偏远渔村的渔民获得更公平的收购价格。沃克表示,技术竞赛的胜负不仅取决于谁先发明,更取决于谁能更好地推广应用。
亚太地区AI应用普及度全球领先。数据显示,新加坡29%的AI用户为“超级用户”,而美国这一比例为12%。亚太地区三分之二的民众对AI持乐观态度,这种乐观情绪是战略优势。在新加坡,教育工作者使用谷歌AI工具备课和定制课程,每周为每位教师节省五小时用于教学辅导。
沃克援引牛津经济研究院报告称,AI未来十年可为全球GDP贡献高达6万亿美元。但这一红利的实现,取决于各国营造适宜的政策环境。
关于知识产权保护,沃克表示,专利制度正在积极应对AI带来的变化。他指出,全球AI专利申请量激增,2024至2025年新增专利数量超过此前十年总和。他强调,专利审查机构需借助AI工具提升审查效率,以甄别大量自动生成的申请。
在版权领域,沃克认为传统版权侧重于“产出”而非“投入”。AI是辅助创作的工具,用户若生成侵权内容,无论使用何种工具,侵权性质不变。法律应聚焦工具的使用方式,同时承认技术的变革性。他呼吁各国采取类似新加坡、日本和欧盟的做法,为AI训练设立明确的数据挖掘例外条款,同时为版权方提供合理的“退出”机制,如通过机器人排除协议和谷歌扩展等工具,让创作者自主选择是否允许其内容用于AI训练。
针对深度伪造等“欺骗性数字复制品”问题,沃克表示,版权法并非最合适的治理工具,此类问题应通过肖像权、反假冒法律解决。谷歌已推出业界领先的SynthID水印工具,并开发了YouTube人脸相似度检测系统。在美国,谷歌支持《2025年禁止假冒法案》和《下架法案》等立法,并愿与各国政府合作,调整既有法律框架以应对新现实。
沃克总结道,过去两百年中,每当变革性技术出现,法律总能灵活调整,适应新挑战。如今面对AI,亦应如此。他强调,AI的“魔力”已经到来,唯有各方携手合作,方能确保其惠及每一个人。
中文翻译:
谷歌在知识产权全球论坛上的发言
编者按:今天,谷歌及Alphabet全球事务总裁肯特·沃克在知识产权全球论坛上发表主题演讲。以下为他的演讲全文,题为“我们实现突破方式的突破:人工智能与创新的未来”。
很荣幸再次来到新加坡,在知识产权周与众多专家相聚于这个变革的关键时刻。
让我们谈谈这一变革的两个方面:
- 第一:人工智能的巨大进步,以及让更多人掌握这些工具的价值。
- 第二:如何在兑现人工智能承诺的同时,保障创作者的合法权益。
让我们开始吧。
谷歌的使命始终是整合全球信息,使人人皆可访问并从中受益。人工智能是我们朝着这一使命前进的巨大飞跃。
我们新的人工智能模型比两年前高效300倍——不是300%,而是300倍。
尽管这是令人难以置信的进步,但更重要的是它们造福人们的途径正在不断扩展。
我们唯一的限制就是我们想象力的广度和深度。
今天的模型不仅能够做出预测,还能独立工作,在不同环境中采取不同行动,并在遇到死胡同时自行纠偏。
这正在带来科学突破——甚至是我们实现突破方式的突破——涵盖医学、能源、材料科学等领域。
例如,牛津大学的研究人员正在使用AlphaFold等人工智能工具——AlphaFold是谷歌开发的一种用于预测蛋白质三维结构的人工智能系统——来设计更好的疟疾疗法,包括一种效果显著提升的新疫苗。
让人工智能惠及每一个人
除了科学领域,从非凡到日常,人工智能正在成为小企业和创业者的力量倍增器,有望改变我们的经济。
ikura,一个日本的人工智能旅游平台,利用谷歌人工智能带领游客走出常规旅游路线。
ikura是一家面向旅行者的日本初创企业,帮助人们体验核心旅游区以外的风土人情,这支撑了那些原本可能接触不到游客的小商家。该人工智能工具消除了旅行中的障碍,如后勤和语言障碍。在印度尼西亚,一家名为Aruna的初创企业正在利用人工智能帮助偏远渔村的渔民为渔获争取更公平的价格。
这样的故事正变得越来越普遍,而这种普及正是让人工智能惠及每一个人的关键。
技术竞赛的赢家未必是最先发明技术的人,而是最能部署技术、将其应用于日常场景并融入经济体系的人。
亚太地区的人们已经拥有全球最高的人工智能采用率。印度和印度尼西亚等国的绝对用户数量最多,而新加坡等地的按人均采用率名列前茅。
新加坡29%的人工智能用户——而美国这一比例为12%——属于我们所说的“超级用户”:即能够委派整个任务、自动化日常流程并使用付费模型的人。他们与人工智能的关系是工作伙伴关系,而不仅仅是搜索引擎的替代品,这意味着他们可以随时调用智能。
乐观是战略优势
高采用率的一个原因在于,亚太地区三分之二的人对人工智能持乐观态度。而这种乐观是将人工智能投入应用的战略优势。
在接受调查的亚太地区人群中,近一半表示他们在工作中使用生成式人工智能每天节省超过一小时。
例如,在新加坡,从小学到初级学院的教育工作者正在使用谷歌教育工作区中的先进人工智能来规划课程和定制教材,为每位教师每周腾出五小时用于教学和学生指导。
新数据显示人工智能可为全球GDP贡献高达6万亿美元
未来几个月,我们将开始看到各类组织和企业不仅在提升效率,还在重塑角色和职能、创造新产品和服务、构建长期竞争优势。
根据牛津经济研究院最近的一份报告,人工智能未来十年可为全球GDP贡献高达6万亿美元。
利害关系从未如此重大——因此我们必须做对。要获得这些收益,取决于各国是否创造了有利的制度环境。
我们最近委托牛津经济研究院撰写了一份报告,剖析版权规则如何影响各国从人工智能中获取的价值。
如何在兑现人工智能承诺的同时保障创作者权益
与此同时,在这次会议上,我们不得不问:“我们能否在兑现人工智能承诺的同时保障创作者权益?”
我认为可以。
让我们看看与人工智能高度相关的专利和版权。
首先,专利。今天的专利制度正在处理重要问题。
谷歌从事人工智能研究多年,拥有规模最大的人工智能专利组合,其中包括一些最基础的人工智能技术专利。
当然,每当出现新技术时,人们就会争相在新的背景下为旧的工作方式申请专利。正如我们在第一代计算机时代所见,随后是互联网时代,我们看到全球人工智能专利申请出现了急剧增长。
而这一次,生成式人工智能工具加剧了这一挑战,因为它使撰写专利申请比以往任何时候都更容易。
因此,根据世界知识产权组织的数据,2024年和2025年发布的新专利比此前十年总和还要多。
我们看到人工智能技术各个层面都出现了权利要求,包括模型架构本身以及这些模型在不同领域的新应用。
当然,并非所有这些权利要求都有效,我们需要确保有相应的工具来评估权利要求的质量,包括利用人工智能分析现有技术并解构由人工智能生成的权利要求。
如果说大语言模型正在把三个要点变成专利申请,那么专利局可能需要用大语言模型把专利申请还原成三个要点。
但从整体来看,专利制度正在支持人工智能工具的开发和部署。
没有必要拆毁一个正在运转的架构,但有必要通力合作以适应这个快速变化的时代。
当我们转向版权时,对话变得更加复杂。一个关键洞见是,传统版权始终关注的是输出,而非输入。
就我们而言,我们在保护知识产权的同时保留创作表达自由的前提下,优先保障人工智能输出的安全性。
我们的方法包括一系列举措,从部署先进的分模态过滤器以防止人工智能模型精确复制训练数据中可能存在的内容,到通过通知—删除机制主动移除侵权内容。
我们相信不必另起炉灶:现有的版权原则是强健的。
人工智能是辅助创作的工具。如果用户创作了侵权输出,无论使用何种技术创作,其侵权性质都不会改变。
无论作品是用铅笔、打字机、个人电脑还是人工智能工具创作的,都不重要。
法律标准保持不变。
正如传统创作工具一样,法律应关注人工智能工具的使用方式,同时承认技术本身的变革性。
关于责任问题,法律始终寻求在工具本身和人们选择如何使用它之间划清清晰界限。
人工智能训练就是学习识别模式
在评估创建人工智能模型所需的训练时,我们同样可以借鉴规范受先前作品启发的文字和图像创作的法律。
人类的创造力始终借鉴前人成果。
如果学生去公共图书馆,阅读书架上的书,学习如何设置情节转折,然后回家写自己的原创小说,他们没有侵犯版权。他们利用这些作品来学习写作技艺,学习文字和段落之间通常如何关联的艺术。生成式人工智能训练的分析方式完全相同,即识别先前内容的模式。
如果法律体制要求开发者对用于训练模型的每一份公开数据都获得商业许可,那将终结人工智能创新。
更好的做法——类似于新加坡、日本和欧盟率先推行的方式——是通过为基于公开可访问内容的训练设立明确的文本和数据挖掘例外条款,为人工智能训练制定清晰规则。印度法院最近也采纳了这一做法。
合理退出机制的重要性
平衡的制度还承认互联网内容的独特规模和性质,并赋予出版商和创作者选择不让其内容用于训练或支撑模型输出的权利。
一个平衡的版权框架,加上明确的文本和数据挖掘例外条款,并不妨碍人工智能开发者与权利人之间就内容访问进行商业谈判——事实上,通过确立清晰的规则,反而促进了这些谈判。
虽然谷歌认为训练模型和通过检索增强生成来提升准确性都属于变革性使用,我们也在与生态系统合作,探索新型伙伴关系和价值交换模式。
我们还实施了各种退出权利,包括通过Google-Extended等控制工具,赋予权利人表达“我选择不参与这个生态系统”的能力;通过robots.txt等长期确立的国际协议,让创作者决定是否允许其内容被用于训练;以及我们更新的搜索控制台协议,让网站所有者管理其链接和内容在生成式人工智能搜索功能中的呈现方式。
广泛的训练和检索增强权利,加上机器可读的退出权,提供了一条合理的中间道路,既能享受前沿人工智能的益处,又能保护版权所有者的权利。
私营和公共部门在应对欺骗性数字复制品方面可以发挥的作用
在结束之前,我还想简要谈谈在知识产权与人工智能的讨论中经常被提及的一个问题:欺骗性数字复制品——即未经授权生成误导性的人工智能深度伪造,模仿个人的声音、面部或肖像。
使用人工智能工具制造这些复制品可能损害当事人的声誉或欺骗受众。
我们认为,行业和监管机构都应在防范欺骗性数字复制品方面发挥作用。
行业有责任建立技术护栏,防止滥用我们的工具并培养信任。
例如,谷歌率先推出了行业领先的SynthID工具,它将不可感知的水印直接嵌入人工智能生成的图像、音频、文本或视频中,降低了关于特定材料由谁创作的欺骗风险。我们还在YouTube上开发了先进的肖像检测工具,扫描系统以识别可能包含创作者面孔的视频。
谷歌的水印工具SynthID是专为人工智能生成内容设计的。它使用户能够识别人工智能生成(或篡改)的内容,有助于促进生成式人工智能的透明度和信任。
在政府方面,我们需要法律来解决这个问题。
版权法不是解决这个问题的合适工具。
版权保护的是原创作品,而非个人身份或事实。因此它不适合解决深度伪造等问题。其他法律体系——针对盗用肖像或虚假代言的法律——更为合适。
这就是为什么在美国,谷歌支持《2025年禁止虚假复制品法案》和《立即删除法案》等法案,以确立针对未经授权和欺骗性数字复制品的明确法律保护。我们也期待与其他政府合作,共同调整现有法律框架以适应这些新现实。
知识产权框架将决定这项非凡新技术的走向
好消息是,我们经过时间考验的法律体系已经在适应人工智能时代的到来。
在过去两个世纪里,每一次变革性技术的出现——从自动钢琴和相机,到广播和电视,再到互联网和人工智能——都有人声称新技术需要在知识产权法领域进行革命。
而每一次,法律都是灵活地应对新挑战,认可新的技术机遇和人类创造力与表达的新媒介。
今天也是如此。
如果我们想确保人们能够充分享受这波非凡新技术带来的好处,就需要保持对话的持续。
人工智能的魔力如今已经触手可及。
通过共同努力,我们可以确保它惠及每一个人。
英文来源:
Google at the Global Forum on Intellectual Property
Editor’s note: Today, Kent Walker, President of Global Affairs for Google and Alphabet, delivered a keynote address at the Global Forum on Intellectual Property. Below is a transcript of his remarks, titled “The breakthrough in how we make breakthroughs: AI and the future of innovation.”
It’s a privilege to be back in Singapore, and here at IP Week with so many experts at a critical time of change.
Let’s talk about two aspects of that change:
- First: The tremendous progress in AI and the value of getting these tools in the hands of more people.
- Second: How we realize the promise of AI while still safeguarding creators’ legitimate rights.
Let’s dive in.
Google’s mission has always been to organize the world’s information and make it universally accessible and useful. AI is a quantum leap in how we make progress against that mission.
Our new AI models are 300 times more efficient than those from just two years ago. Not 300% — 300 times.
And while that’s an incredible advance, what matters even more are the expanding ways they are benefiting people.
We are limited only by the scope and breadth of our imagination.
Today’s models don’t just make predictions, they can work independently, take different actions in different environments, and course correct when they run into dead ends.
That’s leading to scientific breakthroughs — and even breakthroughs in how we make breakthroughs — across medicine, energy, materials science, and more.
For example, today researchers at the University of Oxford are using AI tools like AlphaFold — an AI system developed by Google to predict the 3D structure of proteins — to design improved therapies for malaria, including a much more effective new vaccine.
Making AI work for everybody
And beyond science, turning from the extraordinary to the everyday, AI is a force multiplier for small businesses and entrepreneurs that promises to transform our economies.
ikura, a Japanese AI for travel platform, uses Google AI to take tourists off the beaten path.
ikura, a startup in Japan for travellers, is helping people access experiences outside of core tourist areas, which supports small businesses who might not otherwise reach travelers. The AI tool removes barriers that make travel difficult, like logistical and language barriers. In Indonesia, a startup called Aruna is using AI to help fishermen in remote villages get fairer prices for their catch.
Stories like these are becoming more common, and that diffusion is key in making AI work for everybody.
Technology races are won not necessarily by those who invent the technology first, but by those who deploy it best, putting it to work in everyday settings and integrating it across their economies.
People here in the Asia-Pacific region already have some of the highest levels of AI adoption in the world. Countries like India and Indonesia have the most absolute users, while places like Singapore have some of the highest levels of per capita adoption.
29% of AI users in Singapore — versus 12% of AI users in America — are what we call “super users”: The people who can delegate whole tasks, automate daily routines, and use paid models. Their relationship with AI is a working partnership, not just a search-engine substitute, which means they have intelligence on tap.
Optimism as a strategic advantage
One reason for this high level of adoption is that two-thirds of people in the Asia-Pacific region are optimistic about AI. And that optimism is a strategic advantage in putting AI to work.
Almost half of all people surveyed in APAC say they already save more than an hour a day by using generative AI at work.
For example, here in Singapore, educators from primary schools to junior colleges are using advanced AI in Google Workspace for Education to plan lessons and tailor course material, freeing up five hours per week for each teacher to focus on teaching and mentoring students.
New data details how AI could deliver up to $6 trillion to global GDP
In the coming months, we’re going to start seeing organizations and companies not just making efficiency gains, but reshaping roles and functions, inventing new products and services, and building long-term competitive advantage.
According to a recent Oxford Economics report, AI could deliver up to $6 trillion to global GDP over the next 10 years.
The stakes have never been higher — so we have to get this right. Capturing those gains depends on countries creating the right enabling environment.
We recently commissioned a report by Oxford Economics that unpacks how copyright rules influence how much value countries capture from AI.
How we safeguard creator rights while delivering on the promise of AI
At that same time, and at this conference, we have to ask “Can we safeguard creator rights while still delivering on the promise of AI?”
I think we can.
Let’s look at patents and copyrights, both highly relevant to AI.
First, patents. The patent system today is working through important issues.
Google has been working in AI for many years, and we have the largest AI patent portfolio, with some of the most foundational patents on AI technologies.
Of course, whenever there’s a new technology, people rush to try to patent old ways of working in this new setting. Just as we saw with first computers, and then again with the internet, we’ve seen a dramatic surge in global AI patent applications.
And this time, Generative AI tools have compounded the challenge, by making it easier than ever before to write an application.
As a result, according to the World Intellectual Property Organization, more new patents were published in 2024 and 2025 than in the prior ten years combined.
We’re seeing claims across all aspects of AI technology, including model architectures themselves as well as novel applications of those models in different fields.
Of course not all of those claims may be valid, and we need to ensure we have tools in place to evaluate claim quality, including using AI to analyze prior art and deconstruct claims generated by AI.
If LLMs are turning three bullet points into patent applications, patent offices may need LLMs to turn patent applications back into three bullet points.
But at a high level, the patent system is supporting the development and deployment of AI tools.
There’s no need to tear down an architecture that is working, but there is a need to work together to adapt to a moment of rapid change.
When we turn to copyright, the conversation becomes more complex. One key insight is that traditional copyright has always focused on outputs, not inputs.
For our part, we are prioritizing safeguards for AI outputs as we aim to protect intellectual property while preserving freedom for creative expression.
Our approach includes a range of efforts, from deploying advanced, modality-specific filters that help to prevent AI models from exactly replicating content that might be in the training data, to actively removing infringing material via notice-and-removal systems.
And we believe we don’t have to reinvent the wheel: Existing copyright principles are robust.
AI is a tool that assists creation. If a user creates an infringing output, it's infringing regardless of the technology used to create it.
It doesn’t matter whether a work is created with a pencil or a typewriter, a personal computer or an AI tool.
The legal standard remains.
Just as it does with traditional creative tools, the law should focus on how an AI tool is used, while recognizing the transformative nature of the technology itself.
When it comes to liability, the law has always sought to draw a clear line between the tool and how someone chooses to use it.
AI training is learning to recognize patterns
In evaluating the training necessary to create AI models, we can likewise draw on our laws governing the creation of words and images inspired by prior works.
Human creativity has always drawn on what came before.
If students go to a public library, read the books on the shelves, learn how to create plot twists, and then go home to write their own original novels, they have not infringed copyright. They have used those works to learn the craft, the art of how words and passages typically relate to one another. Generative AI training works in an analytically identical way, recognizing patterns in what’s come before.
A legal regime that required developers to get a commercial license for every piece of publicly available data used to train a model would end AI innovation.
The better approach — similar to those pioneered by Singapore, Japan, and the EU — is to have clear rules around AI training by having clear text and data mining exceptions for training on publicly accessible content. And the courts in India have just followed this approach as well.
The importance of reasonable opt-outs
Balanced regimes also recognize the unique scale and nature of content on the internet, and give publishers and creators the ability to opt out from having their content used to train or ground a model’s output.
A balanced copyright framework, with clear text and data mining exceptions, does not preclude commercial negotiations between AI developers and rights holders for access to content — in fact by establishing clear rules of the road, it facilitates those negotiations.
While Google believes that both training a model and grounding it to improve accuracy are transformative uses, we are also engaging with the ecosystem to explore new types of partnership and value-exchange models.
And we have implemented various rights to opt out, including through controls like Google-Extended, which gives rights holders the ability to say, “I choose not to participate in this ecosystem;” Long-established international protocols like robots.txt, which allow creators to decide whether they want their content to be used for training; and our Updated Search Console protocols, which let website owners manage how their links and content appear in generative AI Search features.
Having a broad right to train and ground, coupled with a machine-readable right to opt-out, offers the reasonable middle way, allowing the benefits of cutting-edge AI while protecting the rights of copyright holders.
The role the private and public sectors can play in responding to deceptive digital replicas
Before I close, I want to also touch briefly on an issue that is often raised in conversations around intellectual property and AI: deceptive digital replicas — the unauthorized generation of misleading AI deepfakes of an individual’s voice, face, or likeness.
Using an AI tool to create these replicas could harm the reputation of the person or deceive the audience.
We believe that both industry and regulators have a role to play in safeguarding against deceptive digital replicas.
Industry has a responsibility to build technical guardrails that prevent misuse of our tools and foster trust.
For example Google pioneered the industry-leading SynthID tool, which embeds imperceptible watermarks directly into AI-generated images, audio, text, or video, reducing the risk of deception about who created a particular material. We also developed advanced Likeness detection tools on YouTube, scanning our system to identify videos that potentially contain the face of creators.
Google’s watermarking tool, SynthID, is designed specifically for AI-generated content. It empowers users to identify AI-generated (or altered) content, helping to foster transparency and trust in generative AI.
And on the government side, we need laws that address this problem.
Copyright law isn’t the right tool to address this issue.
Copyright protects original creative works, not personal identity or facts. So it’s ill-suited for addressing issues like deepfakes. Other sets of laws — those against misappropriating images or fake endorsements — are a better fit.
That’s why in the U.S., Google supports bills like the NO FAKES Act of 2025 and the TAKE IT DOWN Act to establish clear legal protections against unauthorized and deceptive digital replicas. And we look forward to working with other governments on similar efforts to tailor existing legal frameworks to these new realities.
IP frameworks will determine the trajectory of this incredible new technology
The good news is that our time-tested legal systems are already evolving to meet the AI moment.
Every time a transformative technology has emerged over the past two centuries — from player pianos and cameras, to radio and television, to the internet to AI — some people have claimed that the new technology required a revolution in IP law.
And each time, the law instead flexed to address the new challenges, recognizing new technological opportunities and new mediums for human creativity and expression.
The same is true today.
If we want to ensure people can fully tap into the benefits of this remarkable new wave of technology, then we need to keep the conversation going.
The magic of AI is available now.
And by working together, we can make sure that it benefits everyone.