European Commission On Thursday (July 10), it said it released the final version of a voluntary framework designed to help artificial intelligence companies comply with the European Union AI ACT.
General AI Practical Code I'll ask To clarify the legal obligations of providers of general-purpose AI models chatgpt, especially Those Pose Systematic risks Like them that Help con man Develop Chemical and biological weapons.
The code publication “marks important steps to make the cutting edge AI models available in Europe not only innovative, but also safe and transparent.” Henna WilkunenExecutive Vice President of Technical Sovereignty, Security and Democracy in the EU CommissionThe executive arm of statement.
The code was developed by 13 independent experts after hearings from 1,000 stakeholders, including AI developers, industrial organizations, academics, civil society organizations and representatives of EU countries, according to Thursday (10th July). press release. Global Observers Public institutions I participated again.
EU AI Law It was Approved In 2024, it is the first comprehensive legal framework to manage AI. The AI systems used in the EU are safe and secure, Respect for transparency and fundamental human rights.
This law classifies AI applications Risk Category – Unacceptable, high, limited, minimal, and comply with obligations. AI companies that provide services It is being used According to the EU, residents must follow the law. fine It could reach 7% of the world's annual revenue.
The code is voluntary, but signing on AI model companies will benefit from lower management burdens and legal certainty, according to the committee. The next step is for the 27 EU member states and committees to support it.
Read again: The European Commission says it will not delay the implementation of AI laws
Within the Code of Practice
code Structured In three core chapters: Transparency; Copyright; Safety and security.
The Transparency chapter includes a model document form, described by the committee as a “user-friendly” tool, to help businesses demonstrate compliance with their transparency requirements.
The Copyright chapter provides “practical solutions to meet the obligations of AI law to implement policies to comply with EU copyright law.”
The Safety and Security chapter covers cutting-edge systems with systematic risks, outlines “state-of-the-art concrete practices for managing systemic risks.”
The drafting process began with the plenary session in September 2024, with multiple working group meetings, virtual draft rounds and provider workshops.
The code will be effective on August 2nd, but the committee AI Office The rules will be implemented on the new AI model in one year and the existing model in two years.
spokesman for Openai I said Wall Street Journal The company has checked the code and decided whether to sign it or not. a Google The spokesman said the company will also review the code.
Subscribe daily for all PYMNTS AI coverage AI Newsletter.
read more:
