GPT-5, New EU AI Law, and Government Agreements
2 September 2025
Although a summer month, August was full of new releases, battles and regulations updates. But we’ll start with the AI adoption in game development.
A recent Google Cloud and Harris Poll survey shows that 87% of game developers across the U.S., South Korea, Norway, Finland, and Sweden are using AI agents to automate repetitive tasks and streamline workflows. 44% apply AI for content optimization spanning text, code, audio, and video—94% anticipate it will reduce development costs, yet about 25% struggle to quantify ROI. Despite optimism, concerns persist over job displacement and IP issues, following a turbulent period of over 10,000 layoffs and industry strikes over AI policies.
Europe – Regulation, M\&A, Legal Battles & Market Reaction
The EU’s new AI Act entered a critical phase—obliging providers of general-purpose AI (GPAI) models placed on the market from that date to comply with transparency, safety, and copyright requirements. Providers of high-impact models (over 10^25 FLOP) must also notify the Commission and ensure security. Enforcement will begin in 2026, while models predating August 2025 must comply by 2027.
To guide compliance, the EU published a voluntary Code of Practice for general-purpose AI, promoting transparency, data sourcing integrity, safety, and security. Developed with input from over 1,000 stakeholders and approved in July, the code offers firms legal clarity and potential relief from administrative burdens, despite some views that it oversteps legal boundaries.
Reactions split: Meta declined to sign the voluntary Code of Practice, citing legal uncertainty and excessive regulatory overreach, echoing calls from Airbus and Philips for a two‑year delay. In contrast, Google agreed to sign, demonstrating willingness to cooperate with EU regulators.
German AI hardware firm ParTec AG filed a patent infringement suit against Nvidia in Munich, aiming to block DGX supercomputer sales across 18 European countries. The dispute centers on ParTec’s dynamic Modular System Architecture (dMSA), which Nvidia allegedly used without licensing following collaborative exchanges in 2019. A ruling in favor of ParTec could shift AI hardware dynamics in Europe.
Latin America – City AI, Regional Momentum
The Rio de Janeiro City Hall announced partnerships with Oracle and Nvidia to develop the “Rio AI City” initiative, aiming to transform the city into a global AI hub. Nvidia will supply GPUs for municipal models, while Oracle supports infrastructure through Elea Data Centers.A consortium of 12 Latin American countries, led by Chile’s CENIA, is launching Latam‑GPT, a custom AI language model tailored to the region. Set to debut in September 2025, this initiative is supported by over 30 institutions seeking localized, regionally relevant AI.