Startups

OpenAI Hires Transformer Co-Inventor and Ex-White House AI Aide

Noam Shazeer from Google DeepMind and Dean Ball from Trump's AI policy team join ahead of the company's planned IPO.

Omega Editorial· June 19, 2026· 3 min read

OpenAI has secured two significant hires as it prepares for its initial public offering: Noam Shazeer, a foundational figure in modern AI research, and Dean Ball, a former Trump administration AI policy official.

Shazeer announced his departure from Google DeepMind on Wednesday after more than two decades at the company, interrupted only by a three-year period when he co-founded Character AI, an AI role-playing startup. Google brought him back two years ago in a $2.7 billion deal that granted the tech giant access to Character AI's technology, according to TechCrunch, which first reported the moves.

The researcher co-authored the landmark 2017 paper "Attention Is All You Need," which introduced the Transformer architecture that underpins virtually all modern large language models. He most recently served as a co-lead on Google's Gemini project.

Policy reinforcement

OpenAI is simultaneously strengthening its policy apparatus by bringing Ball aboard. The former White House official helped publish America's AI Action Plan during his brief 2025 stint in the Trump administration before returning to the Foundation for American Innovation, a techno-libertarian think tank.

Ball will join OpenAI on July 6 as leader of a new Strategic Futures team, reporting directly to Chief Strategy Officer Jason Kwon. In a blog post, Ball described the unit's mandate as helping company leadership shape frontier AI policy across catastrophic risk, recursive self-improvement, labor market impact, and the relationship between frontier labs, governments, and society.

The team will handle both public-facing policy work and internal governance. Ball emphasized that AI labs will necessarily lead on AI governance decisions, writing that "internal governance will be more central to the future of AI than most people realize."

Why it matters

The dual hires signal OpenAI's strategy to combine technical firepower with political sophistication as it moves toward public markets. Ball's appointment is particularly notable given the current regulatory environment: Anthropic, a key OpenAI rival, recently faced an export control ban from the Trump administration on its latest models, forcing the company to take them down entirely. Ball's White House connections position OpenAI to navigate—and potentially shape—the regulatory landscape that will govern frontier AI development. Meanwhile, securing one of the architects of the Transformer represents a major talent coup in the ongoing competition among top AI labs.

Controversies and context

Shazeer's move comes amid reported internal tensions at Google, where management deleted his posts on internal messaging boards regarding transgender identity and Israel's war in Gaza, according to The Information. Whether those controversies will affect his tenure at OpenAI remains unclear.

The hiring continues a pattern of movement among elite AI researchers across Google, OpenAI, Anthropic, and Meta as companies compete for top talent in the race to develop advanced AI systems.

TechCrunch first reported the details of both appointments.

#openai#noam shazeer#dean ball#ai policy#transformer architecture#ipo

This is an original analysis by the Omega editorial team. Source reporting: AI Watch.

Want systems like this working for your business?

Book a Call

More in Startups

Startups· 3 min read

Snap Spins Off AI Video Team as Dotmo to Cut Costs

The social media company will retain equity and technology ties while offloading expensive generative AI development work.

Via AI Watch · Jun 19, 2026
Startups· 3 min read

AI Model Commoditization Is Here, Says Investor Who Found Facebook

Goodwater Capital's Chi-Hua Chien argues the biggest AI winners will be application companies, not infrastructure providers selling models.

Via AI Watch · Jun 17, 2026
Startups· 4 min read

AI Startup Aaru Predicts Behavior by Training on Actions, Not Words

The company's agents simulate populations using purchase data and outcomes, not traditional surveys—and it tells clients not to trust the results until they test them.

Via AI Watch · Jun 17, 2026