欧盟正大力资助AI安全研究项目，涵盖AI风险建模、问责制、自动驾驶、机器人技术等多个领域，总金额达数百万欧元。然而，AI安全领域的专家和组织却鲜少申请这些项目，导致大量资金流向缺乏AI专业知识的传统咨询公司。文章指出，AI安全社区拥有丰富的专业知识和经验，应积极参与竞标，以获取资源、建立影响力，从而塑造欧洲乃至全球的AI发展方向。文章以一个价值900万欧元的AI安全测试项目为例，详细介绍了其细分领域和申请要求，并提供了寻求免费帮助的途径，鼓励AI安全社区抓住机遇，在AI安全领域发挥关键作用。

💡 欧盟提供大量AI安全研究资金，AI社区应积极争取：欧盟正在资助一系列AI安全相关的研究项目，金额从50万至900万欧元不等，为AI安全领域的专家和组织提供了难得的资源和发展机会。这些项目涵盖了AI风险建模、AI问责制、自动驾驶、机器人技术、信息安全等多个关键领域，目标是推动AI技术的安全和负责任发展。

🚀 传统咨询公司主导项目，AI社区专业优势未被充分利用：目前，许多欧盟AI安全项目被缺乏AI专业知识的传统咨询公司中标，因为他们更熟悉政府采购的官僚流程。这导致了AI安全社区积累十多年的专业知识和经验未能有效应用于这些重要项目中，错失了影响AI发展方向的宝贵机会。

💰 参与投标可获得高额合同与行业影响力：通过成功申请欧盟AI安全项目，组织不仅能获得可观的合同收入（平均50万至900万欧元），还能积累项目经验、建立行业网络，并在塑造欧洲及全球AI政策和技术标准方面发挥重要作用。文章鼓励组织，特别是规模较小的初创企业，考虑与合作伙伴联合投标，以增强竞争力。

🌐 详细拆解900万欧元AI安全测试项目，提供申请指导：文章重点介绍了欧洲委员会一项价值900万欧元的AI安全测试项目，该项目细分为CBRN风险、网络攻击风险、失控风险、有害操控风险、社会技术风险以及Agentic评估接口等六个子领域。文章还提供了关于投标资格、组织规模要求以及如何获取免费的申请支持和咨询的详细信息，旨在帮助AI安全社区克服申请障碍。

Published on July 20, 2025 1:30 AM GMT

Thanks to @Manuel Allgaier of AI Safety Berlin for his suggestion to write this post and his helpful feedback. And thanks to LW/AI Alignment Moderator Oliver for looking over the post. 

LessWrong and AI safety have a unique opportunity: The EU is funding important projects to research AI safety, ranging from AI risk modelling, AI accountability, driverless transport, robotics, to information security and many other topics.

Here’s the problem: We’re not applying.

I work in EU procurement support and I routinely see new AI tenders worth millions [see below] go to the same standard consultants. Most have no idea about the issues involved and I doubt they care much. They know the bureaucratic skills and paperwork inside out so they win by default.

What they lack is AI expertise, which our community has accumulated over more than a decade. I and my EU procurement company will gladly help (for free) anybody in the LW community bridge the application experience gap. But it will only happen if some of you out there take action.

I don’t think the AI safety battle is lost but it does require that the most qualified people and organisations win these projects, get resources, and gain the network and influence to shape AI in Europe (and the world). This isn’t just an opportunity to put our best resources into tackling AI issues but it’s also a chance for your organisations to win big contracts ranging on average from € 500k-€9M+ and lasting a minimum of a year.

 

Case example: EU AI Safety tender worth €9M

To illustrate the opportunities we are missing I want to share this recent tender with you.

This tender comes from the European Commission and is worth €9M and would be perfect for our experts. The purpose of the tender is to test the safety of the large AI models in 6 areas:

CBRN Risks (1.8M€)
Risks of AI models helping in making or using chemical, biological, radiological, or nuclear weapons.Cyber Offense Risks (2M€)
Risks of AI being used to create or scale cyberattacks, find security holes, or generate hacking tools.Loss of Control Risks (1.8M€)
Risks from AI systems becoming too autonomous, deceptive, or hard to control or align with human values.Harmful Manipulation Risks (1.2M€)
Risks of AI manipulating people (e.g. through persuasive conversations or deceptive advice), especially to cause harm.Sociotechnical Risks (1.2M€)
Broader social risks such as bias, discrimination, threats to freedom of speech, or health that can emerge at scale.Agentic Evaluation Interface (1.08M€)
A software system to test how AI models act as agents—making decisions, interacting with environments (e.g. browsers or operating systems), and handling complex tasks.

 

Each of the 6 areas are sub-tenders you can bid for individually, so you don’t have to bid for the whole tender. You can find it here:

https://ted.europa.eu/en/notice/-/detail/272332-2025

 

What are the criteria?

I have worked in the public tender world for over a year, and talked with many organisations from startups to corporations. I have seen many small startups fail because they didn’t understand the public tender game. So I will try to be realistic about what you need to win these projects:

If your org is <15 people you need to apply with a partner organisation
If your org is 15-50 people: Might be able to apply alone.>50 people & >3 years old: Apply by yourselfExperience: Your organisation or institute needs at least one significant reference project. Publications on AI will also help.Organisation type: You can be a private business, startup, SME, corporation, institute, non-profit, or a university.

Got questions? Get in contact

We can advise you if you are fit for a particular tender project and help you with the bid submission too. So if you decide to apply you won’t be doing it alone.

Unsure? Questions? I could organize an Q&A webinar session for interested parties in the LW, EA & AI Safety community. AI Safety Berlin offered to host it. I would present some of the up-and-coming AI tenders and you are free to ask any questions about tenders. I’m willing to help your organisation for free because I think this will help strengthen the AI Safety movement. If you’d join such a webinar, please express interest in the comments or via email to u>Connect@Tendery.ai</u. I’d organize the Q&A session if at least 3-4 people express interest (probably at 5/6pm CET, let me know if that time does not work for you)

Otherwise, write your questions to me directly at Connect@Tendery.ai and be sure to write AI Safety in the subject line and let me know if you want to participate in the webinar.

 

For more info, see my Linkedin.com/in/arustamian/ and www.tendery.ai/en

Discuss