The boards taking shape inside private AI are writing the governance playbook the rest of the market will inherit.

The boards taking shape inside private AI are writing the governance playbook the rest of the market will inherit. Some for how deliberately they are built, some for the questions they raise. The 10 I am watching in 2026:

Anthropic: an independent Long-Term Benefit Trust now appoints a majority of the board. The seat that tipped it was Novartis CEO Vas Narasimhan, in April 2026. The clearest test yet of mission over shareholder control.
OpenAI: a nonprofit foundation controls the for-profit and appoints every director, with a board-level safety committee chaired by Bret Taylor. The most scrutinised structure in tech.
Databricks: co-founder Ion Stoica as executive chairman, and an IPO-grade board that just ruled out going public in 2026. Governance built for a listing it keeps delaying.
Anduril: a defense-AI board chaired by Trae Stephens, who is also a partner at Founders Fund, Anduril's lead backer. National-security oversight is the board's actual job, and the investor-director question sits at the top of it.
Cohere: ex-Google CFO Patrick Pichette on the board, appointed by Inovia Capital, one of its lead investors. The sovereign-AI bellwether outside the US, governed almost entirely by the funds that back it.
Mistral: Europe's flagship, with strategic investor ASML holding a seat at the committee table.
Scale AI: Meta owns 49% and deliberately took none of the votes, a structure built to sidestep control and antitrust questions. Meanwhile its founder sits on Scale's board while running AI at Meta. The sharpest conflicts case here.
Thinking Machines Lab: founder supermajority control, with weighted votes that hand the founder the board. The purest founder-control case here.
Sierra: an interlocking directorate in plain sight. Its CEO, Bret Taylor, also chairs OpenAI.
Safe Superintelligence: founder-locked and deliberately opaque, with no disclosed independent board. It turned down an acquisition anyway.

The pattern: the strongest AI boards are being engineered on purpose, with trusts, safety committees, and weighted votes, while two-thirds of boards report limited to no knowledge or experience with AI (Deloitte, 2025).

Boardroom takeaway. The governance questions these ten are answering now are the ones your board and your portfolio will face next.

It is why I am building BoardLens, the intelligence edge across boards and portfolios. boardlens.ai

Everyone's watching what AI labs build. Almost no one's watching how they're governed.

Enjoyed this? Get the next one.