Who signed the EU’s New AI Code of Practice?
Introduction
The European Union has taken a significant step towards regulating artificial intelligence with the introduction of its voluntary Code of Practice for General-Purpose AI. This code, a precursor to the full implementation of the landmark AI Act, aims to guide companies in developing and deploying AI responsibly, focusing on transparency, safety, and copyright. In a show of commitment, a diverse group of tech companies, from established giants to innovative startups, have signed on, signaling their intent to align with the EU’s vision for a trustworthy AI ecosystem.
The Code of Practice, published on July 10, 2025, provides a framework for companies to demonstrate compliance with the upcoming AI Act. Adherence to the code is voluntary, but it offers signatories a degree of legal certainty and a streamlined path to compliance. The code is structured into three main chapters: Transparency, Copyright, and Safety and Security. While the first two chapters apply to all general-purpose AI providers, the Safety and Security chapter is specifically for providers of the most advanced models that could pose systemic risks.
The development of the Code was a collaborative process facilitated by the European AI Office, which appointed independent experts to lead several working groups. The Transparency and Copyright chapters were shaped by a working group co-chaired by Nuria Oliver, Director of the ELLIS Alicante Foundation, and Alexander Peukert, a leading expert in European copyright law. The comprehensive Safety and Security chapter was informed by the work of three distinct groups: the working group on risk identification and assessment, chaired by Matthias Samwald of the Medical University of Vienna; the group on technical risk mitigation, led by Turing Award winner Yoshua Bengio; and the group on internal risk management and governance, chaired by former MEP and AI governance expert Marietje Schaake.
What does signing the code mean?
For companies that sign the Code of Practice, the primary benefit is a clearer and more predictable path to complying with the AI Act. By voluntarily adhering to the code, these companies can demonstrate their commitment to the Act’s principles, which can result in reduced administrative burdens and greater legal certainty.
Essentially, the code acts as a compliance tool; signatories will likely face less regulatory scrutiny from the EU’s AI Office, which will focus its enforcement on monitoring adherence to the code. This proactive approach allows companies to align with the EU’s expectations early, potentially mitigating risks and fostering trust with regulators and the public.
For more information: EU publishes Code of Practice for General-Purpose AI (GPAI) | by Law and Ethics in Tech | Jul, 2025 | Medium
The signatories
Here are the signatories of the Code of Practice and a look at how they are contributing to the field of general-purpose AI as of August 15, 2025:
Accexible: Headquartered in Barcelona, Spain, Accexible is a health-tech company that uses AI to analyze speech for the early detection of mental and cognitive health conditions like Alzheimer’s and depression. Their models analyze what a person says and how they say it to identify vocal biomarkers associated with these diseases.
AI Alignment Solutions: This signatory represents a field dedicated to ensuring that AI systems align with human values and goals. Companies in this space are developing solutions to make AI safer and more reliable.
Aleph Alpha: This German startup, headquartered in Heidelberg, develops large language models (LLMs) with a focus on transparency and explainability, aiming to provide European alternatives to major US AI products. They develop multilingual AI models for enterprise and government applications.
Almawave: An Italian company headquartered in Rome, Almawave specializes in AI, natural language processing, and big data services. They provide AI-powered solutions that understand text and voice in over 40 languages to optimize customer experience and business processes.
Amazon: The Seattle-based tech giant is deeply involved in AI, from its consumer-facing products like Alexa to its cloud-based AI services on Amazon Web Services (AWS) and the development of its own foundation models.
Anthropic: An American AI safety and research company based in San Francisco, Anthropic focuses on building reliable, interpretable, and steerable AI systems. Their main product is a family of large language models named Claude, developed with a “Constitutional AI” approach to ensure it is helpful, harmless, and honest.
Bria AI: Headquartered in Tel Aviv, Israel, with a presence in New York, Bria provides a platform for visual generative AI designed for commercial use. It offers solutions that address legal complexities and privacy concerns, allowing businesses to create customized visual content.
Cohere: Based in Toronto, Canada, Cohere specializes in developing large language models and AI products for enterprise use, particularly in regulated industries. They provide businesses with powerful and secure AI technology accessible through an API.
Cyber Institute (CyberPeace Institute): Headquartered in Geneva, Switzerland, this non-profit organization works to reduce the harm from cyberattacks on vulnerable communities. They advocate against the malicious use of AI and promote its responsible development to protect international peace and security.
Domyn (formerly iGenius): This Milan-based company specializes in developing “Responsible AI” for regulated industries like finance and government. They provide customizable large language models and AI agents that enterprises can run on their own infrastructure to ensure data privacy and governance.
Dweve: Headquartered in Arnhem, Netherlands, Dweve is building a complete AI ecosystem with the goal of breaking Europe’s dependency on foreign technology. Their platform includes a binary compute engine and autonomous agent frameworks on a decentralized network, designed to deliver faster, more efficient, and secure AI with a focus on data sovereignty and compliance.
Euc Inovação Portugal: With offices in Lisbon and Porto, this Portuguese company focuses on innovation and intellectual property strategies for European and international markets. They are involved in developing European projects and emphasize the use of “digital twin” technology, a virtual representation of a physical object or system, which is a key application of AI for simulation and analysis.
Fastweb: An Italian telecommunications company, Fastweb is experimenting with AI to enhance customer interaction. They have launched FastwebAI, a virtual assistant powered by OpenAI’s technology, to help users explore their service offerings.
Google: A global leader in AI headquartered in Mountain View, California, Google’s work in general-purpose AI is extensive, powering everything from its search engine and translation services to its advanced AI models like Gemini.
Humane Technology: (Information not readily available)
IBM: A long-standing multinational technology company headquartered in Armonk, New York, IBM is heavily invested in AI, offering a wide range of AI-powered solutions and services for businesses through its Watson platform.
Lawise: A legal-tech startup based in Zurich, Switzerland, Lawise aims to democratize legal advice through AI. Their platform provides affordable and accessible AI-powered legal support, with answers validated by legal experts.
Microsoft: The Redmond, Washington-based software giant is a major force in the AI landscape. Microsoft is developing and deploying general-purpose AI across its product ecosystem, including its significant partnership with OpenAI and its own Copilot services.
Mistral AI: A French startup headquartered in Paris, Mistral AI has quickly become a key player in developing open-weight large language models. They champion an open-source approach to make generative AI more accessible and customizable for developers and enterprises.
Open Hippo: Based in Kissing, Germany, Open Hippo is an IT services company that hosts cutting-edge, open-source large language models tailored for business applications. By using independently owned and operated hardware in Germany, they ensure privacy and compliance with GDPR and the EU AI Act, delivering fast and reliable Generative AI solutions.
OpenAI: The San Francisco-based research and deployment company is at the forefront of developing highly capable general-purpose AI models. It is best known for creating ChatGPT, which has brought generative AI into the mainstream.
Pleias: A French private AI lab, Pleias is focused on training the next generation of Language Models specifically for document processing. The company is committed to open science and has coordinated the release of some of the largest open-source corpora for pre-training AI models.
Re-inventa: Based in Madrid, Spain, Re-inventa is an IT services and consulting firm specializing in technology for Contact Centers. They use their proprietary AI solution, #reauditIA, which leverages Natural Language Processing and Large Language Models to automate quality control and analyze customer conversations from both voice and text to improve business quality and optimize resources.
ServiceNow: Headquartered in Santa Clara, California, ServiceNow is a cloud computing company that uses AI to automate and optimize business workflows. Their “AI platform for business transformation” helps organizations streamline IT service management, customer service, and other enterprise operations.
Virtuo Turing: (Information not readily available)
WRITER: Based in San Francisco, this company provides a full-stack generative AI platform for enterprises. It helps businesses create content that is consistent with their brand and style, offering tools to build, activate, and supervise AI agents for various business processes.
XAI: Founded by Elon Musk, xAI is focused on developing advanced AI systems. The company has signed only the Safety and Security Chapter of the Code, indicating a specific focus on mitigating the risks associated with the most powerful AI models.
What about companies that don’t sign?
Companies that choose not to sign the Code of Practice are not exempt from the rules. They will still be fully subject to the obligations of the AI Act once it comes into effect. However, they will have to demonstrate their compliance through other means, without the benefit of the streamlined process offered by the code. This could mean a more intensive and less predictable regulatory process, with a higher likelihood of information requests and direct scrutiny from the AI Office to prove they are meeting the law’s requirements. While not legally binding, the Code of Practice sets a clear benchmark, and non-signatories will need to ensure their internal processes are robust enough to meet the same legal standards.
A notable signatory is xAI, which has committed only to the Safety and Security chapter. This partial commitment reflects a growing trend of specialized focus within the AI industry, where companies may concentrate their efforts on specific aspects of AI development and regulation. On the other hand, Meta stated that it won’t sign the Code of Practice.
Disclaimer: The views expressed in this article are solely my own and do not reflect the opinions, beliefs, or positions of my employer. Any opinions or information provided in this article are based on my personal experiences and perspectives. Readers are encouraged to form their own opinions and seek additional information as needed.**
