Google to Sign EU AI Code of Practice - Latest News

Google to Adhere to EU AI Code of Practice

Google has officially announced its intention to sign the European Union’s general purpose AI code of practice. This voluntary framework is designed to assist AI developers in establishing the necessary processes and systems for adherence to the EU’s forthcoming AI Act.

Meta's Decision to Not Sign

In contrast, Meta declared earlier in the month that it would not be signing the code. The company cited concerns regarding the EU’s approach to AI legislation, characterizing it as an overextension of regulatory power and suggesting a misdirected course for AI development within Europe.

Impending AI Regulations

Google’s commitment arrives shortly before new regulations targeting providers of “general-purpose AI models with systemic risk” take effect on August 2nd. These rules will impact prominent companies like Anthropic, Google, Meta, and OpenAI, alongside other significant generative model developers.

Affected organizations will be granted a two-year period to achieve full compliance with the stipulations of the AI Act.

Google's Reservations

Kent Walker, president of global affairs at Google, acknowledged in a blog post on Wednesday that the finalized version of the code of practice represents an improvement over the EU’s initial proposal. However, he also expressed ongoing reservations concerning both the AI Act and the code itself.

“We continue to be concerned that the AI Act and Code could potentially hinder the advancement and implementation of AI technologies within Europe,” Walker stated.

He specifically highlighted potential issues such as deviations from established EU copyright law, delays in approval processes, and requirements that could compromise trade secrets, all of which could negatively impact European AI development and competitiveness.

Guidelines for AI Companies

By endorsing the EU’s code of practice, AI companies commit to a series of guidelines. These include:

• Providing comprehensive and current documentation regarding their AI tools and services.
• Refraining from utilizing pirated content in the training of AI models.
• Complying with requests from copyright holders to exclude their works from AI datasets.

The EU AI Act: A Risk-Based Approach

The EU’s AI Act, a landmark regulation for AI applications, adopts a risk-based approach. It prohibits certain “unacceptable risk” applications, such as cognitive behavioral manipulation and social scoring.

Furthermore, the act defines specific “high-risk” applications, encompassing areas like biometrics, facial recognition, and the deployment of AI in sectors such as education and employment.

Developers will also be required to register their AI systems and fulfill obligations related to risk and quality management.