Anthropic overtook OpenAI in business adoption and nearly hit a $1 trillion valuation: while Google rolls out personal agents and India launches sovereign LLMs.
Anthropic raises $65B, hits $965B valuation, and overtakes OpenAI in business AI
Anthropic closed a $65 billion Series H round at a $965 billion valuation, making it the most valuable AI company in Silicon Valley, surpassing OpenAI for the first time. The round was led by Altimeter, Dragoneer, Greenoaks, and Sequoia. Alongside the funding news, Anthropic reported a $47 billion annual revenue run rate (up from $10B last year) and released Claude Opus 4.8. Business adoption data from Ramp's May index shows Anthropic now leads OpenAI 34.4% to 32.3%: a first.
Why it matters: For years, OpenAI was the undisputed leader. That is no longer the case. When enterprise adoption flips, product decisions, pricing, and what gets built next all follow. This is a structural shift, not just a funding headline.
Claude Mythos Preview: Anthropic's most powerful model is now in limited hands
Alongside the funding announcement, Anthropic unveiled Claude Mythos Preview, a model with advanced cybersecurity capabilities currently available only to a select group of companies. Mythos is positioned as Anthropic's most capable model yet, purpose-built for high-stakes professional use cases including vulnerability research and security analysis.
Why it matters: AI models purpose-built for security are a different category, they can find and fix vulnerabilities at a speed no human team can match. The fact that access is tightly controlled tells you how seriously the risks are being taken.
Google's I/O tools land: Gemini 3.5 Flash is live, Spark arrives next week
Following Google I/O, several announced products are now actually rolling out. Gemini 3.5 Flash (which rivals larger flagship models at fast speeds) is generally available in the Gemini API and AI Studio. Gemini Spark, a personal agent that takes actions on your behalf across your digital life, launches next week for Google AI Ultra subscribers in the US. Google's Daily Brief feature, which reads your Gmail and Calendar and surfaces a personalised digest each morning, is live today for Google One subscribers.
Why it matters: Announcements are easy. Rollouts are what matter. The week after I/O is when you find out what Google actually shipped. This week, several things are real.
OpenAI's GPT-5.3-Codex: The most capable agentic coding model yet
OpenAI released GPT-5.3-Codex, combining the Codex and GPT-5 training stacks into a single model optimised for agentic coding tasks. It is approximately 25% faster than its predecessor and sets new highs on coding benchmarks. The model is designed for multi-step programming tasks where an AI agent plans, writes, tests, and debugs across a full project, not just autocomplete.
Why it matters: Agentic coding models are starting to handle whole features end-to-end, not just individual lines. Developers who learn to direct these tools will operate at a different scale to those who do not.
Sarvam AI launches India's first government-backed sovereign LLMs
Sarvam AI, chosen by the Indian government under the IndiaAI Mission to build India's first homegrown sovereign language model, has unveiled two models: Sarvam-30B and Sarvam-105B. Both are designed with Indian languages and use cases at their core. The models were announced at the India AI Impact Summit and represent the largest government-backed LLM effort in the Global South to date.
Why it matters: Most AI models are built for English-first, Western-first contexts. A sovereign LLM built for India's languages, legal context, and population scale is a foundational step, not just a technical one. If it works, it changes what AI-powered services in India can actually do.
EU deepfake labeling deadline: 63 days away and platforms are scrambling
From August 2, 2026, the EU AI Act requires all AI-generated content to be marked in a machine-readable format and disclosed clearly to users at first interaction. This applies to providers and deployers operating in the EU. Major platforms are now racing to implement compliant labeling systems, with OpenAI, Google (SynthID), Kakao, and ElevenLabs among those working on compatible content authentication tools.
Why it matters: The EU's move will set a global template. Once large platforms build AI-content labeling infrastructure for EU compliance, it typically rolls out globally. This is the beginning of a world where you will be able to tell (or at least check), whether content was AI-made.
The race to $1 trillion, and what it means when AI companies get that big
Anthropic is at $965B. OpenAI's valuation is being pressured from both below (cheaper models) and beside (Anthropic's adoption lead). The first AI company to cross a trillion-dollar valuation will not just be a milestone, it will shift how governments regulate, how enterprises negotiate, and how much leverage these companies have over the tools millions of people use every day. The pace of that race just accelerated significantly.
Meta Muse Spark debuts + Q1 2026 VC hits $300B record + SynthID goes wider
Meta unveiled Muse Spark, its first flagship LLM built under Chief AI Officer Alexandr Wang's Superintelligence Labs, strong on multimodal reasoning at a fraction of Llama 4's compute cost. Separately, Q1 2026 shattered every VC record: $300B invested globally, with $242B going to AI startups: 80% of all venture capital. And Google announced SynthID adoption by OpenAI, Kakao, and ElevenLabs, expanding its AI content watermarking standard beyond Google's own products.