Transformer Co-Inventor Noam Shazeer Defects from Google to OpenAI
In what may be the most significant AI talent move of the year, Noam Shazeer — co-author of the 2017 paper “Attention Is All You Need” that introduced the Transformer architecture underpinning virtually every modern large language model — announced on June 18 that he is leaving Google DeepMind to join OpenAI as Lead for AI Architecture Research.
The move is remarkable given Shazeer's history: Google reportedly paid $2.7 billion in 2024 to bring him back from Character.AI, where he had been CEO. At Google, he served as vice president of engineering and co-led the Gemini model family. Now, barely two years later, he's heading to OpenAI — a decision that underscores the intense competition for elite AI research talent.
OpenAI also hired Dean Ball, former White House AI policy official, in the same week. Both hires are widely seen as part of OpenAI's push to assemble a world-class roster ahead of its anticipated IPO filing, which the company submitted confidentially to the SEC on June 8.
OpenAI Files for IPO at $850B+ Valuation
OpenAI formally filed a confidential IPO application with the U.S. Securities and Exchange Commission on June 8, 2026, targeting a public listing in Q4 2026 at a valuation between $852 billion and $1 trillion. If successful, it would be one of the largest tech IPOs in history.
The numbers are staggering on both sides of the ledger. First-quarter 2026 revenue reached approximately $5.7 billion, with a full-year target of $30 billion — up from $6 billion in 2024. But profitability remains distant: internal projections suggest losses of $14 billion in 2026 alone, with the adjusted operating margin sitting at -122% in Q1.
The company is bolstering its leadership team with high-profile hires like Shazeer and Ball, and has been holding informal talks with Wall Street banks to prepare for the listing.
Anthropic Fable 5 Export Ban Saga Hits Refund Deadline
The ongoing saga surrounding Anthropic's Claude Fable 5 reached another milestone on June 20, as the refund processing cutoff arrived for customers who had paid for Fable 5 usage credits. The free-trial window for paid subscribers officially closes on June 22.
To recap the extraordinary sequence of events: Anthropic launched Fable 5 — a publicly available, safety-tuned version of its Mythos-class frontier model — on June 9. The model was jailbroken the very next day. On June 12, the U.S. Commerce Secretary issued an export control directive forcing Anthropic to suspend global access to both Fable 5 and Mythos 5 for any foreign national — the first known use of export control authorities to regulate a specific AI model on national security grounds.
New details have since emerged about the chain of events. The White House identified SK Telecom, South Korea's largest carrier and a $100 million Anthropic investor, as a potential Chinese security risk with access to Mythos 5. Separately, Amazon researchers flagged Fable 5 vulnerabilities. White House tech adviser David Sacks reportedly gave Anthropic an ultimatum: fix the jailbreak or de-deploy. CEO Dario Amodei rejected both options, leading to the full export control directive.
Anthropic has pushed back publicly, arguing that if a narrow jailbreak were sufficient grounds to recall a model deployed to hundreds of millions of users, “it would essentially halt all new model deployments for all frontier model providers.” Meanwhile, the company opened its Seoul office on June 17-18 and signed major partnerships with NAVER, Samsung SDS, TCS, and DXC Technology.
Google Launches First Smart Speaker in Six Years with Gemini Built In
Google is shipping the new Google Home Speaker on June 25 for $99.99 — its first audio hardware in six years, and the first built from the ground up with Gemini AI as its core intelligence rather than Google Assistant.
The speaker uses Gemini for Home, enabling natural multi-step conversations, better follow-up question handling, and more intuitive smart home control. The hardware features a rounded fabric-covered design with a light ring that responds to AI activity, a driver twice as large as the Nest Mini, and 2.5x stronger bass. It comes in four colors: Porcelain, Hazel, Berry, and Jade.
Google took its time: over 3.5 million users joined the early access program for Gemini on older speakers, providing feedback that led to more than 2,500 bug fixes before the company committed to new hardware. The approach signals Google's confidence that Gemini-powered voice AI is finally ready for the mainstream consumer market.
Bipartisan “Great American AI Act” Proposes First Federal AI Framework
Representatives Jay Obernolte (R-CA) and Lori Trahan (D-MA) released a 269-page bipartisan discussion draft of the Great American Artificial Intelligence Act of 2026 (GAAIA) on June 4, marking the most serious Congressional attempt at comprehensive federal AI governance.
The bill spans four titles: Frontier AI Governance, Workforce, Cybersecurity, and Research, Development & International Cooperation. Key provisions include mandatory third-party audits for frontier AI models through designated Independent Verification Organizations, whistleblower protections, and a novel requirement that companies disclose AI's role in mass layoffs under the WARN Act.
Perhaps most consequentially, the GAAIA would preempt state AI laws for three years — a provision that could override the patchwork of state regulations including California's AI Transparency Act and Colorado's delayed consumer protection law (now pushed to January 2027). The draft has drawn mixed reactions, with some viewing it as a much-needed federal standard and others dismissing it as unlikely to pass in its current form. Either way, it establishes the parameters for what federal AI regulation might look like.