A week prior to the implementation of new regulations concerning general purpose artificial intelligence (GPAI) in the European Union (EU), major developments have emerged regarding the industry’s response to the EU’s voluntary Code of Practice on GPAI. This initiative, aimed at tools like ChatGPT and Gemini, is designed to enhance transparency, security, and copyright protections in AI technologies. Notably, Meta, the US-based tech giant, has publicly declared its refusal to sign the Code, criticizing it for potentially hindering innovation in the rapidly evolving field of artificial intelligence.
The EU’s Code of Practice, recently released by the European Commission, outlines a voluntary framework intended to help GPAI providers navigate compliance with the AI Act. Companies that choose to sign the Code are expected to align with the AI Act, thereby gaining a clearer legal framework, while those that opt out face increased scrutiny. Among the organizations committing to the Code is Anthropic, a US AI firm known for its Claude AI assistant, which has emphasized that the Code embodies critical principles like transparency and safety. OpenAI has also announced its intent to sign, viewing the moment as an opportunity for European innovation and progress.
The process of drafting the Code has faced considerable criticism from various stakeholders, particularly rightsholders concerned about potential copyright infringements and tech giants worried about reluctance to innovate under the new regulations. Microsoft, represented by President Brad Smith, has indicated a likelihood of signing as well, aiming to project a more constructive stance amid geopolitical uncertainties. In contrast, Meta stands firm in its decision against signing, with the company’s Chief Global Affairs Officer suggesting that the Code introduces legal ambiguities that exceed the requirements of the AI Act.
Experts, such as Gry Hasselbalch, have pointed out that even companies like Meta, which choose not to sign the Code, are still obliged to comply with the AI Act. As such, signing becomes more of a symbolic act than a substantive one; it does not exempt companies from adhering to the AI regulations set to take effect soon. Additionally, Hasselbalch argues that the current AI landscape reflects a geopolitical struggle, with the AI Act having been crafted through a democratic process aimed at providing certainty for developers and users in the EU market.
The broader implications of the new Code extend beyond individual corporate decisions, resonating within the context of international relations and regulatory frameworks. The AI Act’s stipulations, classifying AI systems based on their associated risks, are critical to understanding how these technologies will be regulated moving forward. Observations point towards a lack of clarity in the roles and responsibilities that various AI providers will hold, potentially complicating the landscape as providers adapt to new compliance measures.
As the August 1 deadline approaches for companies with existing GPAI models, the EU intends to release a list of signatories, further cementing the commitment to establishing a structured framework for AI. Given the mixed responses from major players in the tech industry, it remains to be seen how compliance will unfold and what tangible impacts these decisions may have on innovation, competition, and the overall evolution of artificial intelligence within the EU.
