%20(1).png)
The European Commission recently published a list of companies that have signed its AI code of practice, a voluntary framework for providers of general-purpose AI models to comply with upcoming EU regulations. Major U.S.-based players such as Amazon, Google, Microsoft, IBM, OpenAI, and Anthropic signed on, along with European firms Aleph Alpha and Mistral AI, while Elon Musk’s xAI committed only partially. Dozens of smaller companies also pledged compliance. The code outlines requirements around transparency, copyright, and security, and serves as a pathway to meet obligations under the EU’s landmark AI Act, which officially took effect on August 2.
Meta, however, has emerged as the most notable holdout. The company argued that the code imposes legal uncertainty and exceeds the scope of the AI Act, and further warned that it could hinder the development of advanced models in Europe. Joel Kaplan, Meta’s global affairs chief, criticized the EU approach as regulatory overreach that could stifle innovation and limit opportunities for European startups. Yet despite pushback from several global tech firms, the Commission has stood firm, publishing implementation guidelines and setting compliance deadlines for systemic-risk AI providers, with stricter obligations applying by 2027.
Find more details Politico and TechCrunch.
The summer of 2025 marked a defining moment in the AI race, with Anthropic launching Claude Opus 4.1 and OpenAI releasing ChatGPT-5 within 48 hours. Both models are meant to push performance boundaries but with different targets.
ChatGPT-5 is designed to simplify AI use with a unified model that adapts automatically to task complexity, integrates natively with Gmail and Calendar, and introduces built-in personalities for personalized interactions. According to OpenAI, it significantly reduces hallucinations compared to previous versions and excels in rapid front-end development and productivity-focused tasks, albeit at a high premium cost.
Claude Opus 4.1, meanwhile, positions itself as the “gold standard” for technical work. With increased debugging accuracy, extended 64K-token reasoning, and measurable ROI for developers, Anthropic says it delivers unmatched precision in complex code analysis and enterprise-level problem-solving.
Explore how both models compare to each other and to ChatGPT-4 at Digidop.
Google has published the most detailed breakdown yet of the environmental footprint of its Gemini AI system, revealing per-prompt averages of 0.24 watt-hours of electricity, 0.03 grams of CO₂ emissions, and 0.26 milliliters of water — about as much as a few seconds of TV viewing or five drops of water. These figures cover not just AI chips but also supporting CPUs, memory, idle backup systems, and data center overhead, with custom TPUs (Google’s proprietary equivalent to GPUs) accounting for only 58% of the total energy. The disclosure highlights how efficiency improvements have cut Gemini’s per-query energy use by 33 times in just a year, though overall emissions continue to rise due to soaring AI demand.
The report also underscores the tension between transparency and greenwashing. While praised for providing unprecedented insight, researchers note gaps such as the absence of total query volumes and questions about how Google calculates emissions from its clean energy purchases. With data centers consuming over 30 million megawatt-hours annually, double what they did in 2020, Google is exploring a mix of renewables, demand-response agreements, and even advanced nuclear partnerships to keep AI growth sustainable. Experts say this disclosure is an important milestone but stress the need for standardized reporting across the industry to fully understand AI’s environmental toll. fifty-five recently released an in-depth study on the cost of Generative AI with the BrandTech Group and Scope3 – find it here along with our GenAI carbon calculator.
Find more details at CarbonCredits and MIT Technology Review.
U.S. District Judge Amit Mehta has ruled that Google will not be forced to sell its Chrome browser or Android operating system as part of remedies for its illegal monopoly in online search, a decision that sparked a rally in Alphabet’s shares. While rejecting the Justice Department’s more aggressive proposals, the court ordered Google to share a one-time set of valuable search data with qualified competitors and barred it from signing exclusive deals that would block rivals’ distribution. The ruling also allows Google to continue billions in revenue-sharing payments to partners like Apple and Mozilla, which critics say perpetuates its dominance, but Mehta argued banning the practice could destabilize the wider ecosystem.
The decision marks the most significant U.S. antitrust remedy against a tech giant in 25 years, though it fell short of what regulators sought. Critics, including DuckDuckGo and advocacy groups, called the measures weak and warned that Google will still be able to stifle competition, including in AI search. Mehta defended his narrower approach, citing the rapid rise of AI challengers like OpenAI and Perplexity as evidence that the competitive landscape is already shifting. Both the Justice Department and Google are weighing appeals, with observers expecting the case could ultimately reach the Supreme Court. Meanwhile, Google continues to face other legal battles over its app store and advertising technology monopolies, underscoring an intensifying scrutiny of Big Tech.
More info at Reuters and The Verge.
The European Union has paused its long-anticipated antitrust fine against Google’s AdTech operations, with several reports linking the delay to fears of disrupting sensitive trade negotiations with Washington. European regulators had internally targeted Sept. 1 for announcing penalties and forcing changes to Google’s business model, but concerns grew that the Trump administration might retaliate by unraveling progress on a transatlantic trade deal and reimposing tariffs. The potential fine, expected to be smaller than Google’s record €4.3 billion penalty in 2018, was shaped by the EU’s new antitrust chief Teresa Ribera, who favors reforming practices over imposing heavy financial punishments.
Behind the scenes, the decision has exposed rifts within the European Commission, with Ribera reportedly at odds with trade commissioner Maros Sefcovic. Key critics, including Germany’s Monopolies Commission, warned that competition policy risks becoming a bargaining chip in international politics. Accused of favoring its own AdX exchange to entrench its dominance, Google maintains that its integrated approach mirrors industry norms.
Find more at PYMNTS and AdWeek.
Discover all the latest news, articles, webinar replays and fifty-five events in our monthly newsletter, Tea O'Clock.
.svg)
