AI News Roundup — May 11, 2026
Anthropic hits a $30B revenue run rate after 80x growth, Claude Opus 4.7 launches with JPMorgan, GPT-5.5 becomes the default for all ChatGPT users, and the US government starts pre-screening AI models before launch.
AI News Roundup: May 11, 2026
Anthropic had the biggest week in AI company history. GPT-5.5 is now the default for every ChatGPT user. The US government is now pre-screening frontier AI models before launch. And the enterprise AI race is accelerating faster than anyone predicted.
Anthropic's Biggest Week Ever
Anthropic dropped six major announcements in five days last week, and the AI industry is still processing them.
Claude Opus 4.7 Launches With Financial Services Focus
Anthropic released Claude Opus 4.7, its most capable model yet, with particular strength in advanced software engineering and complex multi-step reasoning. Opus 4.7 represents a meaningful step up from 4.6, especially on the class of problems where previous models would hallucinate or give up. We've been building with the Claude Agent SDK ourselves — if you're curious how autonomous AI agents work in practice, our AI agents and Bitcoin wallets deep dive covers the emerging agentic economy.
Jamie Dimon's Name on the Press Release
The headline move: Anthropic launched ten pre-built AI agents for financial services with JPMorgan CEO Jamie Dimon as the named launch partner. The agents cover wealth management, risk analysis, compliance, and trading infrastructure — built specifically for the world's largest banks.
This is not a pilot. This is Anthropic planting its flag in enterprise finance while OpenAI has been pursuing the same sector aggressively with its operator model. Having Dimon's name attached is significant: JPMorgan has historically been slow to embrace external AI partnerships.
Claude Agent SDK Opens to All Developers
The Claude Agent SDK — previously in restricted beta — is now open to every external developer. Any developer can now build autonomous agents on top of Claude with full access to tool use, memory, and multi-step task execution.
Revenue: 80x Year-Over-Year — and a $30B Run Rate
Anthropic's Q1 2026 revenue grew 80x year-over-year, pushing the company to a $30 billion annualised revenue run rate — up from roughly $9 billion at end of 2025. CEO Dario Amodei told CNBC the growth was "too hard to handle," with the company renting data center capacity from Elon Musk's xAI just to keep up with demand. This is the steepest single-quarter revenue jump any frontier AI company has publicly disclosed — and it's 8x what the company had planned for.
The revenue trajectory tells the story: $87M run rate in January 2024 → $1B by December 2024 → $9B end of 2025 → $30B today. The AI infrastructure demand driving these numbers is the same force reshaping power grids and data centers worldwide.
Claude Mythos Preview: The Security Model
Anthropic also previewed Claude Mythos, a specialized model that excels at identifying software vulnerabilities and security flaws. In pre-release testing, Mythos identified thousands of previously unknown zero-day vulnerabilities across every major operating system and browser — flaws that had survived decades of human review. It reproduced vulnerabilities and developed working exploits on the first attempt in over 83% of cases.
CEO Dario Amodei met with senior Trump administration officials at the White House to demonstrate Mythos's capabilities. Rather than a public release, Anthropic launched Project Glasswing — a coalition of ~40 technology companies including AWS, Apple, Microsoft, Google, CrowdStrike, and Palo Alto Networks — with controlled access to the model for defensive security work.
The security angle connects directly to the quantum threat landscape — systems that can find classical software weaknesses will be critical infrastructure as post-quantum migration accelerates.
GPT-5.5 Is Now the Default for Every ChatGPT User
OpenAI shipped GPT-5.5 Instant as the new default model across all ChatGPT tiers on May 5. Previously available only in the API since its April 24 launch, GPT-5.5 Instant is now the "chat-latest" model for all users — free, Plus, and Pro alike.
GPT-5.5 is a significant leap: it scores 82.7% on Terminal-Bench 2.0 and 58.6% on SWE-Bench Pro (real-world GitHub issue resolution) — both state-of-the-art results. It excels at code, online research, data analysis, and operating software autonomously until a task is complete. The full GPT-5.5 and GPT-5.5 Pro remain available in the API for higher-intensity workloads.
Notably, VentureBeat reported that GPT-5.5 narrowly beats Claude Mythos Preview on Terminal-Bench 2.0, setting up a genuine head-to-head benchmark race between the two labs.
US Government Will Now Pre-Screen AI Models Before Launch
In a significant policy development, the Center for AI Standards and Innovation (CASI) announced formal agreements with Google DeepMind, Microsoft, and Elon Musk's xAI allowing US government agencies to evaluate AI models before public release.
The Trump administration's AI oversight framework — emphasising safety testing without imposing strict regulations — is taking shape through these voluntary agreements. The goal is to identify national security risks, dual-use concerns, and capability thresholds before models reach the public. This represents a meaningful shift: the government is no longer reacting to AI releases but getting ahead of them.
Notably absent from the initial announcement: Anthropic — who appears to be operating through a separate White House channel via the Mythos demonstrations, rather than the CASI framework.
Google Gemini Gets Agentic: Webhooks and Multimodal RAG
Google's Gemini API received two significant updates for enterprise agentic workloads this week:
Event-Driven Webhooks allow long-running tasks to notify users upon completion without polling — a quality-of-life improvement that matters enormously for production agentic systems where tasks may take minutes or hours. This closes a meaningful gap with Claude's tool use framework.
Multimodal RAG now natively handles documents, images, and video in a single knowledge pipeline. Gemini 3.1 Pro ($2.00 input / $12.00 output per million tokens) is positioning itself as the go-to choice for multimodal enterprise search. The broader DeFi and crypto analytics space is already exploring AI-assisted yield strategy analysis using exactly these capabilities.
The Multi-Model Routing Trend Is Accelerating
A new generation of developer infrastructure is emerging: multi-model routing, where production applications automatically route different tasks to the most cost-effective model capable of handling them.
Industry benchmarks show optimised multi-model routing reduces total AI API costs by 60–80% compared to routing all traffic through a single premium model. As GPT-5.5, Claude Opus 4.7, Gemini 3.1 Pro, and Grok 3 occupy different price/capability niches, developers who build smart routing logic gain a structural cost advantage. Tools like LiteLLM, Portkey, and OpenRouter are seeing rapid adoption as the infrastructure layer for this approach.
Bitcoin miners transitioning to AI infrastructure are at the centre of this shift — see our piece on how miners are becoming AI infrastructure companies for the full picture.
What to Watch This Week
The Anthropic flywheel: With the Agent SDK open, expect a wave of Claude-powered fintech and legal applications in the coming weeks. Watch for enterprise announcements as partners build on the financial agents framework.
OpenAI operator push: GPT-5.5 Instant as the default ChatGPT model is a land-grab for the consumer interface. Watch for new operator partnerships to counter Anthropic's JPMorgan move.
Claude Mythos and government procurement: If Mythos gets formal government adoption, Anthropic's $30B run rate trajectory could accelerate further still.
Sources current as of Monday May 11, 2026. Previously: AI Roundup May 7.