March 30, 2026 – AI Daily Recap

From Jensen Huang declaring AGI has arrived to Anthropic’s accidental model leak and Apple opening Siri to rival AI assistants, the weekend’s news underscored just how fast every layer of the AI stack is shifting — models, chips, platforms, and policy alike.

Jensen Huang Says “We Have Achieved AGI”

In a wide-ranging conversation with Lex Fridman released over the weekend, NVIDIA CEO Jensen Huang declared that AGI has been achieved. “I think it’s now. I think we’ve achieved AGI,” Huang told Fridman, pointing to AI systems that can already build billion-dollar web services — even if those businesses may not last forever. The statement immediately ignited fierce debate. Critics questioned Huang’s narrow definition, while supporters framed it as a pragmatic acknowledgment of frontier-model capabilities. Either way, the claim from the CEO of the world’s most valuable chipmaker carries weight — and keeps the AGI timeline debate very much alive.

Anthropic’s “Mythos” Model Exposed in Accidental Data Leak

An unsecured database containing nearly 3,000 unpublished Anthropic documents revealed the existence of “Claude Mythos” (internally codenamed “Capybara”), a new model tier positioned above Opus that Anthropic describes as “a step change” in capabilities. Draft blog posts indicate Mythos is “dramatically higher” in performance on coding, reasoning, and cybersecurity tasks — with the company warning it is “currently far ahead of any other AI model in cyber capabilities.” The leak, discovered by researchers at LayerX Security and Cambridge, sent cybersecurity stocks down 3–7%. Anthropic confirmed the model is in early access testing with select customers and removed public access after being notified by Fortune.

Apple Will Open Siri to Gemini, Claude, and Rival AI Assistants

Apple is preparing a major Siri overhaul in iOS 27 that will allow competing AI services — including Google Gemini and Anthropic’s Claude — to function as alternative assistants on iPhone, iPad, and Mac. A new “Extensions” architecture will let users route queries through their preferred AI service from a dedicated settings menu, moving Apple well beyond its existing ChatGPT-only integration. The company is also developing its own chatbot, reportedly codenamed “Campos” and powered by Google’s Gemini technology. More details are expected at WWDC on June 8.

Arm Ships Its First Proprietary Chip in 35 Years

In a historic shift, Arm unveiled the AGI CPU, a 136-core, 3-nanometer data center processor designed specifically for AI inference — the company’s first-ever in-house chip after 35 years of licensing its architecture to others. Meta is the launch customer, with OpenAI, Cloudflare, and SAP among seven additional committed buyers. CEO Rene Haas projected the chip line will generate $15 billion in revenue by 2031, roughly six times Arm’s entire 2025 annual revenue. Up to 64 CPUs — around 8,700 cores — fit in a single air-cooled rack. Arm’s stock jumped 16% on the news.

Harvey AI Hits $11 Billion Valuation as Legal AI Booms

Legal AI startup Harvey closed a $200 million round at an $11 billion valuation, co-led by GIC and Sequoia — the latter’s third consecutive lead investment in the company. The raise caps a 3.5x valuation jump in roughly a year (from $3 billion in February 2025 to $5 billion, then $8 billion, and now $11 billion). Founded in 2022, Harvey’s AI tools for contract analysis, compliance, and litigation are now used by more than 100,000 lawyers across 1,300 organizations, signaling that vertical AI applications are attracting serious capital.

Weekend Roundup: More Stories Worth Watching

A few more developments that crossed the wire this weekend. Waymo doubled its weekly paid rides to 500,000, reaching halfway to its annual target in a single year of operation. Google launched Gemini 3.1 Flash Live and Search Live across 200+ countries, alongside tools that let users import chat history from competing platforms. Mistral shipped Voxtral TTS, a 3-billion-parameter open-weight text-to-speech model that outperformed ElevenLabs in blind testing. And on the policy side, US Senators Sanders and AOC introduced legislation proposing a moratorium on new AI data center construction pending comprehensive federal regulation.

The Big Picture

This weekend’s news paints a picture of an AI industry where the frontier is moving on every axis simultaneously. The models keep getting more powerful (Mythos), the chips to run them are being reimagined (Arm AGI CPU), the platforms are opening up to let users pick their own AI (Apple + Siri), and the money keeps flowing into vertical applications that deliver measurable ROI (Harvey). Meanwhile, Jensen Huang’s AGI declaration — agree with it or not — is a signal that the industry’s own leaders believe we’ve crossed a threshold. The real question now isn’t whether AI can do remarkable things, but how quickly businesses, governments, and society can adapt to the pace of change.

Sources

• Yahoo Finance — Nvidia CEO Jensen Huang claims AGI has been ‘achieved’
• Sherwood News — Jensen Huang: We have achieved AGI now… sort of
• Fortune — Anthropic ‘Mythos’ AI model revealed in data leak
• MacDailyNews — Apple to open Siri to third-party AI chatbots in iOS 27
• CNBC — Arm launches its own CPU, with Meta as first customer
• TechCrunch — Harvey confirms $11B valuation: Sequoia triples down
• The Neuron — Everything that happened in AI March 28-29, 2026
• Harvey Blog — Harvey raises growth round at $11 billion valuation