Here's the Lowdown on the EU's General Purpose AI Code of Practice
AI Code involvement persists among U.S. corporations, asserts EU authority figure
We're taking a gander at the EU's General Purpose AI (GPAI) Code of Practice, a set of guidelines intended to help providers comply with the EU's AI Act. This code is in the works, and it's crucial for GPAI model providers to get a grip on these rules because they'll be enforceable come August 2025.
The Headache
The EU's AI Office is handling the creation of the Code, consulting with a variety of stakeholders like AI developers, industry associations, and civil society groups to ensure it aligns with the AI Act's requirements. The current draft is being revised to address concerns brought forth in the feedback process.
The Buzz
The Code has been met with a fair share of controversy, particularly regarding its definition and scope. There's discussion about what constitutes a GPAI model and who qualifies as a provider, with preliminary guidelines addressing these issues. Clarifications on concepts like "placing on the market" and open-source exemptions are being discussed.
Another contentious issue is handling systemic risks associated with powerful AI models. Given the ethical and technical challenges in defining and managing these risks, this brings up some challenging questions.
Who's Who
- Naysayers: Unsurprisingly, Big Tech companies and publishers have raised concerns that the Code might violate the EU's Copyright laws.
- The Architects: The EU Commission and the AI Office are the primary architects of the Code, shaping the regulatory framework for GPAI.
- The Influencers: Big Tech companies, like Google, Microsoft, and Facebook, are heavily involved due to their relationship with powerful AI models. Their input and compliance strategies will significantly impact the Code's implementation.
The Future
The plan is to roll out the rules before August 2023, when the rules on GP AI tools, such as large language models like ChatGPT, go into effect. By August, the Commission will decide whether to formalize the Code through an implementing act based on companies' intentions to sign it.
The Report
A recent report by Corporate Europe Observatory (CEO) and LobbyControl suggests that Big Tech companies exerted pressure on the Commission to water down the Code of Practice. However, the Commission asserts that all participants had equal opportunities to engage in the process.
The Path Forward
An alternative option, wherein businesses commit only to specific parts of the Code, has not been proposed yet. But if it ever does, these businesses would need to fulfill their obligations in a different way, according to the official.
Want to Know More?
- AI Act
- EU Policy
- Artificial Intelligence
- The EU Commission and the AI Office, being the primary architects of the General Purpose AI (GPAI) Code of Practice, have been exploring the use of artificial-intelligence (AI) technology to enhance its enforcement, potentially introducing a voluntary WhatsApp group for stakeholders to discuss updates and concerns.
- As the Code of Practice undergoes revisions to address feedback, AI model providers interested in staying informed may consider joining this proposed AI-based communication platform, staying up-to-date on revisions and deadlines, such as the August 2025 enforcement date.
- Despite concerns from Big Tech companies regarding potential copyright law violations, a working group within these companies, comprising experts from Google, Microsoft, and Facebook, is actively engaging in dialogues to help shape the interpretation and voluntary implementation of the GPAI Code of Practice, thereby minimizing systemic risks associated with powerful AI models.
