AI Briefing

The Treasury Secretary and the Fed Chair just summoned every major bank CEO to Washington because an AI model got too good at hacking. Anthropic's Mythos has officially spooked the people who run the financial system. Meanwhile, Meta is throwing another $21 billion at CoreWeave to feed the compute beast, Perplexity quietly hit $450 million ARR after one monster month, and Anthropic launched managed infrastructure that lets you deploy AI agents without babysitting them.

01PolicyBloomberg ↗

Bessent and Powell Summon Bank CEOs to Emergency Meeting Over Anthropic's Mythos Cyber Threat

Treasury Secretary Scott Bessent and Fed Chair Jerome Powell convened an urgent meeting at Treasury headquarters with the CEOs of Citigroup, Bank of America, Goldman Sachs, Morgan Stanley, and Wells Fargo to warn them about the cybersecurity risks posed by Anthropic's Claude Mythos Preview model. The bank chiefs were already in Washington for a Financial Services Forum board meeting when the special session was called. JPMorgan's Jamie Dimon was the only major banking CEO who could not attend. This marks the first time a single AI model release has triggered an emergency response from both the Treasury and the Federal Reserve, signaling that frontier AI capabilities are now being treated as systemic financial risks on par with market shocks or cyberattacks.

02IndustryCNBC ↗

CoreWeave and Meta Expand AI Infrastructure Partnership to $35 Billion Total

CoreWeave announced an expanded agreement with Meta Platforms to provide AI cloud capacity through December 2032 for approximately $21 billion, bringing the total value of their partnership to roughly $35 billion. The new spending runs from 2027 to 2032, layered on top of a prior $14.2 billion arrangement signed just six months earlier. The deal includes some of the first deployments of NVIDIA's next-generation Vera Rubin platform, with a key focus on inference workloads rather than just training. Meta's AI capital expenditure for 2026 is projected at $115-135 billion, nearly double last year, underscoring just how aggressively the company is building out infrastructure to support both its open-source Llama models and the new proprietary Muse series.

03IndustryPYMNTS ↗

Perplexity Hits $450M ARR as AI Agent Pivot Drives 50% Monthly Revenue Surge

Perplexity's annual recurring revenue topped $450 million in March 2026 after a staggering 50% jump in a single month, up from about $305 million before a strategic pricing overhaul. The revenue explosion was driven by the launch of 'Computer,' an AI agent that handles tasks like shopping, email management, and data analysis, alongside a shift from flat subscriptions to usage-based pricing with credits. The company now has over 100 million monthly active users and tens of thousands of enterprise clients paying between $20 and $200 per month. Perplexity has set an internal target of $656 million ARR by year-end, and its trajectory shows that the pivot from AI-powered search to autonomous agents may be the most commercially significant strategic bet in the AI startup space right now.

04ProductSiliconANGLE ↗

Anthropic Launches Claude Managed Agents to Handle Enterprise AI Infrastructure

Anthropic released Claude Managed Agents in public beta on April 8, a hosted service that lets developers deploy autonomous AI agents without managing sandboxing, state management, authentication, or error recovery themselves. Priced at $0.08 per session hour on top of standard token costs, the service includes sandboxed code execution, credential management, checkpointing, and execution tracing. Enterprise early adopters Notion, Rakuten, and Sentry are already in production, with Sentry shipping a debugging agent that autonomously writes patches and opens pull requests in weeks instead of months. This is Anthropic's clearest move yet toward becoming an infrastructure platform, not just a model provider, absorbing the hardest parts of agent deployment so enterprises can focus on business logic.

05ProductMacRumors ↗

ChatGPT Arrives on Apple CarPlay, Becoming the First AI Chatbot for Drivers

OpenAI shipped ChatGPT for Apple CarPlay on April 2, making it the first AI chatbot available in vehicles through Apple's in-car platform. The integration is voice-only by design, with no text or imagery displayed, and requires iOS 26.4 or later. But the bigger story is what Apple did to make it happen: it quietly created an entirely new CarPlay app category called 'voice-based conversational apps,' meaning Gemini, Claude, and any other AI assistant that meets the rules can now ship into the same slot. There's no wake word, so drivers must manually open the app, and ChatGPT cannot control vehicle or phone functions. Still, this establishes a new distribution channel for AI assistants that could become as significant as the smartphone app store was for mobile.