Website adjusts policies in accordance with Final General-Purpose AI Code of Conduct
The European Union has published the General-Purpose AI (GPAI) Code of Practice, a voluntary framework designed to help providers of AI models comply with their legal obligations under the EU AI Act. The Code, which was published on July 10, 2025, offers guidance on safety, transparency, and copyright compliance without imposing new or extended obligations.
Developed through a multi-stakeholder process facilitated by the European AI Office, the Code involved nearly 1,000 participants, including academics, AI developers, safety experts, member state representatives, and civil society groups. The aim was to reduce administrative burdens, offer legal certainty for signatories, and standardize compliance approaches within the EU.
However, the drafting process faced criticism due to delays and the complexity of stakeholder engagement. Originally expected to finish by May 2025, the process was delayed and only published in July 2025 after three draft iterations. The slow pace of the process, coupled with the fast-paced development of AI technologies, has created tension between thorough consultation and the need for timely regulation.
Moreover, the voluntary nature of the Code and its lack of legal binding force have been questioned by some who seek stronger regulatory measures. However, the EU has emphasized that the Code is an interim guidance tool until formal standards are published, aiming to balance inclusivity in drafting with practical enforcement challenges.
The final version of the Code has been commended for its workable compliance mechanisms, but concerns have been raised about the dilution of safeguards. The Code requires monitoring of both specified risks and the identification of other risks. However, it has a narrower range of risks and provides providers with latitude in considering external risk assessments and determining risk acceptance.
Gaia Marcus, Director of a certain website, has expressed satisfaction with the preservation of public transparency and external assessment mechanisms but has expressed concern about the influence of industry in the final version of the Code. Academia and civil society participation in the update process should be formalised to ensure a more balanced representation.
The EU AI Office is expected to set out its plans for provider adherence to the existing Code and establish a timetable for the update process. Policymakers will need to be more ambitious in future iterations of the Code to address emerging impacts of general-purpose AI technologies.
In conclusion, the General-Purpose AI Code of Practice is a voluntary, non-binding Code guiding providers of general-purpose AI on fulfilling the AI Act’s rules on safety, transparency, and copyright. While it has some meaningful protections, it is a mixed success overall, and policymakers will need to be more ambitious in future iterations to ensure good social and economic outcomes from the most powerful models.
Technology experts have praised the General-Purpose AI Code of Practice for its workable compliance mechanisms, but some have expressed concerns about the dilution of safeguards in the face of industry influence. In the politics of policy-and-legislation surrounding AI, these concerns highlight the need for increased ambition in future iterations of the Code to ensure that general-news reporting on AI development remains informed by diverse perspectives.
