EU AI fines are live, OpenAI slashed prices while ChatGPT hit 1 billion users, and Maharashtra launched India's boldest state-level AI policy yet.
EU AI Act is now fully enforceable, fines of up to €35M start today
The EU AI Act reached full enforcement status on August 2, 2026. Providers of large AI models, including GPT-5.6, Claude, and Gemini, had to comply with GPAI obligations from August 2025; the August 2 milestone activates the broader enforcement regime. Fines can reach €35 million or 7% of global turnover. High-risk AI systems (like those used in hiring, credit, and healthcare) get an extended deadline to December 2027, but the core rules are in effect now.
Why it matters: This is no longer a framework in progress, it is live law. If you use AI tools at work that touch hiring, lending, or health decisions, your employer needs to be compliant. For everyday users, it means AI companies must be more transparent about how their products work.
OpenAI cuts prices 80%, and ChatGPT just hit 1 billion weekly users
OpenAI reduced the price of its GPT-5.6 Luna model by 80%, dropping it to just $0.20 per million input tokens: one of the cheapest frontier-model prices ever. Separately, ChatGPT crossed 1 billion weekly active users this week. The same platform that had 100 million users in early 2023 now serves ten times that every week.
Why it matters: Cheaper AI means more developers building tools, more small businesses using AI, and more students experimenting. The price drop accelerates adoption in markets where cost has been the main barrier.
Google Gemini 3.5 replaces Google Assistant on Android
Google has pushed Gemini 3.5 deeper into Android, officially replacing Google Assistant as the default AI on most Android devices. Gemini now handles voice commands, search, and task-based work that Assistant previously managed. This puts a frontier-class AI model in the hands of billions of Android users globally without any additional setup.
Why it matters: Most people will experience AI for the first time not through ChatGPT but through their phone's default assistant. Gemini becoming that default is a quiet but massive distribution event for AI adoption.
Adobe and Johns Hopkins built "Wonder", any image becomes a 3D world
Adobe Research and Johns Hopkins University announced "Wonder," a model that turns a static image or video clip into a persistent 3D world you can navigate in six directions at 16 frames per second. It was announced on July 29, 2026, and is being described as a significant step toward AI-generated immersive environments.
Why it matters: Content creation is about to change. Architects, designers, educators, and game developers could generate navigable 3D environments from a single photo, no 3D modelling software needed. This is early-stage, but the direction is clear.
AI data centers will consume 565 TWh of electricity this year, and energy is now the bottleneck
Gartner reports that global data center electricity consumption will hit 565 terawatt hours in 2026, up 26% from last year. AI-optimised servers alone account for 31% of that demand. In the US, dedicated AI data centers already take one-third of the country's total data center electricity. Analysts say energy availability (not chips or talent) is now the main constraint on how fast AI can scale.
Why it matters: The AI race is increasingly an energy race. Countries and companies that secure clean, reliable power at scale will have a structural advantage in AI development. This is why nuclear, solar, and grid investment is now directly linked to AI strategy.
Maharashtra launches India's most ambitious state-level AI policy
Maharashtra announced its AI Policy 2026, targeting Rs 10,000 crore in investment and 1.5 lakh new jobs by 2031. The policy includes a Maharashtra AI Mission to train 2 lakh youth and professionals, AI cities in five innovation zones, a State AI Data Exchange, a 20% subsidy for MSMEs adopting AI, and a new platform called the Maha AI Tools Hub. It is modelled on the national IndiaAI Mission framework.
Why it matters: Maharashtra is one of India's largest economic engines. A state-level AI policy with this scale of ambition (training targets, MSME subsidies, and dedicated infrastructure) signals that India's AI push is moving from national announcements to local execution.
50+ Global South nations meet in Geneva next week to draft an AI Sovereignty Compact
From August 12 to 14, ministers, researchers, and civil society leaders from more than 50 nations of the Global South will convene at the United Nations in Geneva for the AI Sovereignty Summit. They are drafting the Geneva Compact on AI Sovereignty, a binding framework designed to give developing nations a collective voice and pooled computing power, borrowing from the model of how countries once negotiated vaccine prices together. If agreements are signed, this could be the most significant multilateral AI governance moment of 2026.
Boston makes AI literacy a graduation requirement + Meta wearables get smarter
Boston Public Schools will become the first major US school district to require AI fluency for high school graduation, starting September 2026. Separately, Meta is expanding AI capabilities on its smart glasses with wrist-based sEMG input, letting you control AI with subtle hand gestures rather than voice commands. And the OECD and EU published the AILit Framework in June, defining 19 AI literacy competences for schools globally.