中国作为目前在研新药数量全球第二的经济体，在AI制药领域同样实现了快速发展。在去年召开的中国医药工业发展大会上，一份医药工业数智化转型报告指出，目前人工智能技术已经“较高程度”介入到了创新药生产之中。
• “十四五”期间获批上市创新药达113个，其中已“较高程度”运用到人工智能技术
• 中国、美国先后发布AI制药指导文件，相关领域监管提上日程
2024年诺贝尔化学奖结果带来的震动仍未平息：三位蛋白质结构预测领域的科学家捧得该奖项，其中两位供职于谷歌旗下的人工智能公司DeepMind。他们开发的AI模型AlphaFold对2亿多种蛋白质结构获得了突破性理解，而蛋白质结构是药物研发的重要起点，这意味着AI制药的想象空间被极大拓展了。
中国作为目前在研新药数量全球第二的经济体，在AI制药领域同样实现了快速发展。在去年召开的中国医药工业发展大会上，一份医药工业数智化转型报告指出，目前人工智能技术已经“较高程度”介入到了创新药生产之中。
植德律师事务所合伙人李筠怡律师长期服务于生命科学与医疗健康产业，她在与客户的沟通和对产业的观察中也留意到AI制药——尤其AI辅助药物研发的发展。“其中布局最多的是先导化合物的设计优化合成，其次是化合物筛选和靶点发现。”她说。
药物研发过程中，“从苗头到先导”是一个关键环节，意味着具有初步活性的苗头化合物被优化改善为能够成为候选药物的先导化合物，而AI技术可以有力协助实现这一突破。
不过李律师坦言：“目前AI制药仍处于探索和早期发展阶段，国内外至今没有一款完全利用AI技术发现和设计的新药获批上市。”
实际上，在药物研发之外，AI技术目前在制药的其他关键阶段——包括临床前研究、临床试验、药品生产和销售等，也扮演起一定角色。
其中AI技术运用比较成熟的是药物CMC（Chemistry, Manufacturing, and Controls，化学、制造和控制）全流程管理以及GMP（Good Manufacturing Practice，药品生产质量管理规范）合规应用，“尤其是生产过程的质量集中管控，能实现对工艺改进、场地和设备变更和批次变化的在线监测和追溯”，李律师说。
另外一个正在寻求突破的是临床试验环节。“这是新药研发成本最高的环节，”李律师说，“不过目前AI应用相对有限，原因是临床试验对数据的质与量要求非常高，而大量研发和临床数据为各制药公司积累的非公开数据，保密性高。因此，在缺乏数据的情况下，即使有先进的AI技术，短期内也较难降低临床试验的风险。”
监管积极跟进
在科学仍探索突破的领域，监管反应又如何？由于药物关乎人类的健康福祉，因此监管机关也以不同形式积极跟进AI制药发展。
2024年11月，中国率先发布了《卫生健康行业人工智能应用场景参考指引》。李筠怡律师告诉ALB，其中总结了许多经典的“人工智能+药物研发”场景，例如智能药物研发、智能药物临床试验辅助、智能药品临床综合评价辅助等，“可见监管机关整体呈支持态度”。
无独有偶，今年1月初，美国食品药品管理局（FDA）发布了《使用人工智能支持药品和生物制品监管决策的考量》指南草案，其中也罗列出在药品生命周期使用AI模型的情况。总的来说，FDA认为AI模型的核心作用是生成信息或数据，用来支持药物安全性、有效性、质量的决策，例如减少动物试验的数量、用模型展开临床药物分析、整合数据以提高对疾病的理解、分析数据以评估临床结果、持续监控并评估上市后不良反应等。
有趣的是，FDA并未在草案中涉及目前AI运用较为广泛的药物发现、提升运营效率等环节，因为在其中使用AI并不会影响药物质量或患者安全。
不过，李律师指出，中美上述指引只表明了监管的初步态度，目前对于药品全生命周期的AI模型使用及如何监管，“监管机构尚未形成统一共识和出台明确清晰的法规或指南。”
但在罗列用途之外，FDA草案也明述了AI的挑战，并对未来的监管思路提出设想，李律师认为，FDA作为监管科学决策的引领者，其态度和论述可能为未来监管路径提供方向指引。
例如，FDA指出AI参与制药带来的挑战包括：训练数据的具体情况可能会影响输出结果的可靠性；AI模型往往暗箱运行，需要透明化其方法和流程；AI模型输出的结果会受具体部署环境、引入新数据等影响，因此对模型本身也要进行生命周期维护。
基于此，FDA提出了一种“基于风险的可信度评估框架”思路，帮助申办方评价在特定使用环境中AI模型的可信度。虽然这并非强制性要求，但FDA鼓励申办方尽早与FDA接洽，以展开评估、较早识别潜在风险，李律师介绍道。
李律师补充道，除了监管AI模型，实际上AI工具也将极大赋能药物监管本身。“例如2024年6月中国发布《药品监管人工智能典型应用场景清单》，其中就提出考虑利用AI技术对药品相关资料进行数据挖掘与智能分析，实现对药品监管全生命周期中潜在风险的精准识别等。”她说。
律师提供服务
AI制药涉及的另一重挑战在于知识产权，就此，律师已经在为创新药企提供相关服务。
植德律师事务所合伙人唐华东律师主要服务于医药健康领域的知识产权需求。他指出，AI制药的专利申请存在着天然挑战：一方面AI参与药物研发涉及到大量数据、算法，申请专利相对复杂；另一方面目前专利有效期为20年，而AI算法迭代非常快，制度与现实存在一定的不相匹配。
“此外，创新药企业使用AI辅助制药发明，发明人如何认定？”目前这是一大难点，“根据通用性的法律法规和司法实践，自然人对于整体实验构思、模型开发、数据收集、潜在化合物筛选、实验验证等过程的参与都可能成为考量因素”，唐律师说。
正因专利问题复杂，才特别需要律师提供专业服务。“我们可以协助企业构建全球AI相关专利战略和布局，保护核心算法和药物研发技术的创新性，包括在不同法律体系下申请和保护AI相关的药物专利或算法专利，以及通过专利FTO等手段识别潜在的侵权风险等。”唐律师介绍道。
此外，他认为AI制药模型涉及大量数据，如何明确数据所有权以避免侵犯原始数据提供者的权利将是关键；另外用户难以真实了解AI算法模型内部工作原理，使用模型时发生侵权问题将难以举证，这些都会是未来知识产权律师协助应对的难题。
AI制药及创新药发展方兴未艾，李筠怡律师认为在其他方面律师也有用武之地。“受政策支持，这些年大量优秀的科学家和研发人员回国创业。2023年，中国生物医药授权交易数量居全球第二，仅次于美国，这说明中国创新药资产质量已经很高，能得到跨国药企和海外市场的认可。”
“在产业化方面，我们可以助力AI创新药企业实现商业价值最大化，通过License授权、整体出售或拓展海外市场等多种方式，推动企业安全、高效地实现成果转化。”她说。
此外，“近年医药行业的投融资交易十分活跃，港股18A和科创板制度的推出使得尚处于研发阶段的创新药企有机会通过上市融资。外部律师能够就此提供全方位支持，包括对接投资机构、配合尽职调查、参与交易文件的起草与谈判，助力企业顺利完成融资”。
而在政策监管和经营合规方面，“我们将持续关注监管机构对AI制药企业和AI 技术/模型在制药领域应用的最新监管要求，协助企业评估药物全生命周期的合规风险，审阅各类关键协议，协助建立适用于AI技术和应用场景的合规体系和流程”，李律师说。
Regulators, lawyers keep pace with AI-driven drug development
- During China's 14th Five-Year Plan, 113 innovative drugs were approved, many incorporating AI at an advanced level.
- China and the U.S. have released AI drug development guidelines, bringing regulatory oversight into focus.
The impact of the 2024 Nobel Prize in Chemistry is still reverberating. Three scientists in protein structure prediction won the award, including two from Google’s AI company, DeepMind. Their AI model, AlphaFold, has made groundbreaking progress in predicting over 200 million protein structures, which are crucial for drug development. This has significantly expanded the possibilities of AI-driven pharmaceuticals.
As the world's second-largest market in terms of ongoing new drug projects, China has also seen rapid progress in AI-driven drug development. At last year's China Pharmaceutical Industry Development Conference, a report on the digital transformation of the pharmaceutical industry noted that AI has already been widely integrated into innovative drug production.
Li Junyi, a partner at Merits & Tree specializing in life sciences and healthcare, has observed the rise of AI-driven drug discovery through industry insights and discussions with clients. "The most prominent applications are in lead compound design, optimization, and synthesis, followed by compound screening and target discovery," she says.
However, Li also points out that AI-driven drug discovery is still in an early stage. "So far, no drug fully discovered and designed using AI has been approved for market release, either in China or globally," she states.
Beyond drug discovery, AI is playing a role in other key stages of pharmaceutical development, including preclinical research, clinical trials, manufacturing, and sales. The most mature applications of AI are in CMC (Chemistry, Manufacturing, and Controls) process management and GMP (Good Manufacturing Practice) compliance. Another area seeking breakthroughs is clinical trials. "This is the most expensive stage of new drug development," Li explains, "but AI applications in this field remain relatively limited."
Regulators keeping pace
How are regulatory bodies responding in a field where scientific breakthroughs are still unfolding? Since pharmaceuticals directly impact human health and well-being, regulators are actively following the development of AI-driven drug discovery in various ways.
In November 2024, China became the first country to release the Reference Guide for AI Application Scenarios in the Health Sector. Li explains that this document outlines several key “AI + drug development” scenarios, such as AI-assisted drug discovery, AI-supported clinical trials, and AI-driven comprehensive drug evaluations. “This demonstrates that regulators are generally showing a supportive attitude,” she says.
Similarly, in early January this year, the U.S. Food and Drug Administration (FDA) released a draft guidance titled Considerations for the Use of Artificial Intelligence to Support Regulatory Decision-Making for Drugs and Biological Products, which also lists various AI applications across the drug lifecycle.
However, Li points out that these guidelines from China and the U.S. only represent an initial regulatory stance. “Regulators have not yet reached a unified consensus or established clear and specific regulations on AI model usage throughout the entire drug lifecycle,” she explains.
Beyond simply listing AI applications, the FDA’s draft also highlights challenges associated with AI and proposes regulatory considerations for the future. Li believes that as a leader in science regulation, the FDA’s stance and analysis could serve as a directional guide for future regulatory frameworks.
For instance, the FDA identifies key challenges in AI-driven drug development, including the reliability of model outputs depending on training data specifics, the “black box” nature of AI requiring greater transparency, and the impact of deployment environments and new data on AI model performance—necessitating lifecycle maintenance of the models themselves.
In response, the FDA is proposing a “risk-based credibility assessment framework” to help applicants evaluate AI model reliability within specific use cases. While this is not a mandatory requirement, the FDA is encouraging early engagement with regulators to conduct assessments and identify potential risks as soon as possible, Li explains.
Legal support needed
Another major challenge in AI-driven drug development is intellectual property, and lawyers are already providing related services to innovative pharmaceutical companies.
Tang Huadong, a partner at Merits & Tree specializing in intellectual property in the pharmaceutical and healthcare sectors, highlights the inherent challenges in patent applications for AI-driven drug discovery. “On one hand, AI-driven drug development involves vast amounts of data and algorithms, making patent applications complex. On the other hand, patents are valid for 20 years, while AI algorithms evolve rapidly, creating a mismatch between regulatory frameworks and real-world technological progress,” he explains.
“In addition, when innovative drug companies use AI-assisted drug invention, how should the inventor be defined?” he continues. “According to general legal principles and judicial practice, factors such as a natural person’s involvement in overall experimental design, model development, data collection, potential compound screening, and experimental validation may all be considered.”
Given the complexity of patent-related issues, professional legal services are particularly necessary. “We can assist companies in building a global AI-related patent strategy, protecting core algorithms and drug development innovations, applying for and securing AI-related drug or algorithm patents across different legal systems, and identifying potential infringement risks through patent FTO (freedom-to-operate) analysis,” Tang explains.
With AI-driven drug discovery and innovative pharmaceuticals still in their early stages, lawyers have an important role to play in other areas as well. Li points out that favorable policies have led many top scientists and researchers to return to China to start businesses. “In 2023, China ranked second globally in biotech licensing transactions, just behind the U.S. This reflects the high quality of China’s innovative pharmaceutical assets, which are gaining recognition from multinational pharmaceutical companies and international markets,” she notes.
“In terms of commercialization, we can help AI-driven pharmaceutical companies maximize their business value by facilitating technology licensing, company acquisitions, or international market expansion—ensuring a secure and efficient path for product transformation,” she explains.
Furthermore, “the pharmaceutical industry has seen a surge in investment and financing transactions in recent years. The introduction of the Hong Kong Stock Exchange’s Chapter 18A and the STAR Market listing rules has enabled pre-revenue innovative pharmaceutical companies to secure funding through IPOs. External legal counsel can provide comprehensive support in this process, including connecting with investors, assisting with due diligence, and drafting and negotiating transaction documents to help companies successfully complete financing,” she adds.
On the regulatory front, “we are closely monitoring the latest regulatory requirements for AI-driven pharmaceutical companies and the application of AI technologies and models in drug development. We assist companies in assessing compliance risks throughout the drug lifecycle, reviewing key agreements, and establishing compliance frameworks and processes tailored to AI technologies and application scenarios,” Li states.
