Answer-First Lead
Anthropic confidentially filed its S-1 with the SEC on June 1, setting up what could be the largest AI IPO in history at a ~$965B valuation — beating OpenAI to the public markets. The same 48 hours saw Microsoft’s Build 2026 keynote drop Project Polaris (a homegrown coding model replacing GPT-4 Turbo in GitHub Copilot), Alphabet announce an $80B equity raise with a $10B injection from Berkshire Hathaway’s new CEO, and Florida sue OpenAI and Sam Altman personally, alleging ChatGPT aided mass shooters. The thread connecting all of it: the OpenAI era is fracturing, and everyone is building alternatives.
🔍 THE BOTTOM LINE
Anthropic goes public, Microsoft builds its own models, Alphabet raises $80B with help from Warren Buffett’s successor — three different answers to the same question: what comes after OpenAI? The answer, it turns out, is “anything else.”
📰 Today’s Stories
1. Anthropic Files for IPO — Largest AI Public Offering in History
Anthropic confidentially submitted a draft S-1 registration statement with the SEC on June 1, according to the company’s official announcement. The filing comes just days after its $65B Series H round at a ~$965B post-money valuation led by Altimeter, Dragoneer, Greenoaks, and Sequoia. The confidential filing means revenue figures, underwriter names, and expected float aren’t public yet — but the message is clear: Anthropic is going public before OpenAI.
The S-1 filing caps an extraordinary 12-month capital raise: $30B in April, $65B in June, now an IPO. Dario Amodei has reportedly told investors the company is profitable on an operating basis, a claim that will be tested when the SEC’s confidential review process eventually produces public documents.
Why it matters: This is the moment AI left private capital dependency for public market accountability. OpenAI’s stalled restructuring means Anthropic gets to frame the narrative first. The S&P 500 already rewrote listing rules for AI companies. The question now is whether AI valuations look like hyperscaler multiples or growth software — and this IPO is the test case.
2. Microsoft Build 2026 — Project Polaris Cuts the OpenAI Cord
Microsoft’s Build 2026 keynote (June 2-3, San Francisco) dropped its biggest surprise on Day 1: Project Polaris, a homegrown coding LLM that will replace GPT-4 Turbo as the default engine inside GitHub Copilot. Built with chain-of-thought and tree-of-thought reasoning, Polaris targets multi-file refactoring and directly competes with Claude Code.
The broader message from Satya Nadella’s keynote: Microsoft is building its own AI stack. The MAI model suite spans image, voice, and transcription. Windows becomes an “agent platform.” Copilot moves to token-based billing. A Windows Agent Store opens for third-party AI agents.
Why it matters: Microsoft was OpenAI’s largest investor and exclusive cloud partner. Now it’s building models to not need OpenAI. Project Polaris is the clearest signal yet that the Microsoft-OpenAI partnership is transitioning from exclusive to competitive.
3. Alphabet Raises $80B in Equity — Berkshire Hathaway Puts in $10B
Alphabet announced an $80B equity capital raise on June 1 to fund AI compute infrastructure, according to an SEC filing. The structure includes a $10B private placement from Berkshire Hathaway — the first major investment under Berkshire’s new CEO after Warren Buffett.
The raise is proportionally one of the largest equity offerings in corporate history, creating potential dilution of roughly 5-6% for existing shareholders. The proceeds will fund data centers, TPU clusters, and fiber infrastructure for Google’s AI buildout.
Why it matters: Berkshire Hathaway investing $10B in Alphabet is a generational signal. Warren Buffett famously avoided tech. His successor’s first big bet is AI infrastructure. This isn’t a speculative venture bet — it’s a conviction that Google’s AI compute is a long-duration asset worth owning.
4. Florida Sues OpenAI and Sam Altman Personally — First State-Level AI Product Liability Suit
Florida Attorney General James Uthmeier filed an 83-page civil lawsuit against OpenAI and CEO Sam Altman on June 1, alleging the company knowingly concealed ChatGPT’s risks and that the product “aided and abetted mass shooters” in deadly rampages. The suit follows an April criminal investigation into ChatGPT’s alleged role in a Florida State University shooting.
Altman is named personally — a notable legal escalation that treats executives as directly liable for product safety failures. The complaint alleges OpenAI made false safety claims while deploying models with dangerous capabilities, and that the company knew from internal testing that safety guardrails were insufficient.
Why it matters: This is the first state-level lawsuit treating an AI model like a defective consumer product. If Florida prevails, it creates a template for product liability claims against AI companies across all 50 states. OpenAI has not yet formally responded in court.
5. OpenAI’s Codex Goes After White-Collar Work — Launches Sites and Enterprise Plugins
OpenAI released a major Codex update on June 2 that lets AI agents build interactive enterprise workspaces called “Sites” and use role-specific plugins — effectively extending Codex beyond coding into knowledge work. The company also published a report titled “The Next Era of Knowledge Work,” positioning Codex as a productivity tool for everyone, not just developers.
The update allows non-technical teams to create custom agent workspaces for their specific workflows, with plugins for CRM, project management, document processing, and analytics.
Why it matters: Codex started as a coding tool. This update turns it into an enterprise platform. The timing — landing the same week Microsoft announces Polaris and Anthropic files for IPO — underscores just how contested the enterprise AI market has become.
6. Project Glasswing Expands to 150 New Critical Infrastructure Partners
Anthropic announced it is expanding Project Glasswing — its AI-powered code security initiative — from ~50 initial partners to approximately 150 new organizations across 15+ countries. New partners span power, water, healthcare, communications, and hardware sectors — industries the initial cohort largely didn’t cover.
Partners have already found more than 10,000 high- or critical-severity security flaws using Claude Mythos Preview. The expanded cohort includes critical infrastructure providers and comes with heightened security requirements for participation.
Why it matters: Code security is where AI’s most immediate and least controversial value lives. Finding 10,000 vulnerabilities in weeks is a number no manual security review can match. This isn’t theoretical — it’s deployed, working, and scaling.
7. New Zealand Passes Law Allowing AI to Make Welfare Benefit Decisions
The New Zealand Parliament passed legislation allowing the Ministry of Social Development to use automated decision-making for welfare benefits. Social Development Minister Louise Upston says the change will reduce delays, errors, and unnecessary debt while freeing up staff. Opposition MPs warned of “Robodebt-style” disasters, referencing Australia’s automated debt recovery scandal that caused widespread harm.
Why it matters: NZ becomes one of the first countries to legislate AI-driven welfare decisions. The Robodebt comparison is serious — when automated systems get benefit calculations wrong, real people lose access to food and housing. The safeguards in this law will set a precedent for how NZ approaches AI in public services.
8. Iran-Linked Hackers Using ChatGPT and Gemini for Cyber Operations
The Financial Times reported that Iran’s military and affiliated hacker groups are increasingly using Western AI tools — including ChatGPT and Gemini — for malware development, phishing campaigns, and operational planning. Check Point Research documented related operations under the “Nimbus Manticore” designation. OpenAI and Google have not formally commented on the specific FT reporting.
Why it matters: The same AI tools enterprises use for productivity are being dual-used by state-aligned threat actors. This creates enforcement pressure that didn’t exist before: when your chatbot helps write malware for a hostile government, you can’t claim neutrality indefinitely.