OpenAI confirmed ChatGPT now has roughly one billion weekly active users, a number no consumer software has reached this fast. Alongside the milestone, OpenAI slashed GPT-5.6 Luna pricing by 80% to $0.20 per million input tokens, and introduced ChatGPT Work, an agentic system that executes complex multi-hour projects across team files and apps.
Why it matters: When AI gets a billion users and drops its price at the same time, it stops being a tool people try and becomes infrastructure people depend on. If you haven't built AI into how you work yet, the cost excuse is gone.
Anthropic releases Claude Opus 5, its fourth new model in two months
Anthropic released Claude Opus 5 on July 24, following Mythos 5, Fable 5, and Sonnet 5 in quick succession. The model brings a 1 million-token context window and stronger agentic capabilities, designed for complex reasoning tasks that unfold over many steps. The U.S. Commerce Department now requires government review before Claude Fable 5 can be commercially launched.
Why it matters: Four major model releases in eight weeks is an extraordinary pace. The 1-million-token window means you can feed an entire book, dataset, or codebase into one prompt, that unlocks use cases that simply weren't possible six months ago.
EU AI Act August phase is live: deepfake labelling is now mandatory
The European Commission's August 2026 enforcement phase of the EU Artificial Intelligence Act is now in effect. Platforms must identify and label AI-generated and synthetic content, including deepfakes. Regulators gained new powers to inspect advanced AI models, demand technical records, and act against companies that fail to comply. General-purpose AI model providers face stronger transparency obligations.
Why it matters: This is the moment AI policy shifts from guidance to enforcement. If you create content in Europe or for European audiences, labelling rules apply to you now, not eventually.
US Commerce Dept: the most powerful AI models now need government clearance
The U.S. Commerce Department set national security review gates for frontier models that exceed certain capability thresholds. Major releases including GPT-5.6 and Claude Fable 5 must go through government review before public launch. OpenAI's development of ChatGPT 6 also revealed significant cybersecurity vulnerabilities in testing, adding to pressure on both companies and regulators.
Why it matters: Governments are no longer just watching AI, they are gatekeeping it. This will slow some launches and is likely to shape which countries get frontier AI access first.
Maharashtra launches a 500-crore AI startup fund and 12 incubators
Maharashtra announced its AI Policy 2026, including a dedicated 500-crore-rupee AI Startup Venture Fund and plans to establish 12 AI incubators across the state. This sits within India's broader national AI push: the IndiaAI Mission's $1.25B program now has 38,000+ GPUs provisioned, and a new report places India's AI market opportunity at $126 billion by 2030, and a $1.7 trillion GDP impact by 2035.
Why it matters: State-level AI funding at this scale is new. If you are building an AI product or looking for your first real opportunity in the field, India's infrastructure window is open right now.
Speak hits $100M revenue with 10 million language learners
Speak, an AI-powered conversational language learning app, crossed $100M in annual revenue with around 10 million active users. The app uses AI to simulate real conversations, making immersive practice accessible outside a classroom. Meanwhile, Alice, an AI exam-prep tool described as "Duolingo for exams," is already being used by more than 10,000 college students after its recent pre-seed raise.
Why it matters: Education AI is moving past hype into real revenue at scale. Conversational AI for learning is proving that people will pay (and keep paying) for AI that helps them actually improve at something.
Google's Gemini 3.6 Flash underperformed. Gemini 4 is the response.
Google's Gemini 3.6 Flash underperformed on key benchmarks, raising questions about Google's AI trajectory when competitors are releasing multiple flagship models per quarter. Gemini 4 is in development and positioned as the response, but in an environment where Claude Opus 5 and GPT-5.6 are already shipping, the pressure on Google's next release is unlike anything it has faced before. Watch Gemini 4's launch timing and benchmark performance closely.
Humanoid robots near industrial scale + US deepfake federal remedy expands
Boston Dynamics and Google DeepMind are pushing humanoid robots closer to real industrial deployment. Separately, 48 US states now have some form of deepfake regulation (up from 46 at the start of the year), and the DEFIANCE Act, passed unanimously in January 2026, gives victims of non-consensual AI deepfakes a federal right to sue creators, distributors, and hosting platforms.