Q&A with Alondra Nelson, former deputy director of Biden's OSTP who spearheaded the White House's AI Bill of Rights effort, on the AI policy challenge and more
I'm Ezra Klein. This is “The Ezra Klein Show.” — [MUSIC PLAYING] — It would be easy, listening to the discourse about A.I. …
Context & Ripple Effects
Alondra Nelson built the OSTP AI policy machinery's first major public output: the AI Bill of Rights, a voluntary blueprint issued while she was deputy director, before any formal executive action existed. This interview lands in April 2023, between that blueprint and the harder instruments that followed.
The arc since has been one of escalation: Biden went on to sign an executive order directing NIST, DHS and other agencies to create safety standards, Obama publicly weighed in on the EO and the pattern of incumbents asking for regulation, Deputy Chief of Staff Bruce Reed took charge of the administration's AI strategy, and OMB converted principle into paperwork with mandatory annual AI reports across federal agencies.
First-order effects
- Nelson's platform here lets the AI Bill of Rights' chief architect argue its case just as the administration debates whether voluntary principles need to harden into enforceable rules — shaping how the eventual executive order gets framed.
- OSTP's role in AI policy, established under founding National AI Initiative Office director Lynne Parker and carried by Nelson, becomes a template successor directors like Arati Prabhakar inherit.
Second-order effects
- A voluntary-first framework lowers the barrier for large incumbent labs to shape compliance norms they can absorb, the dynamic Obama flagged when he noted incumbents lobbying for regulation — smaller entrants face the same rulebook without the same compliance staff.
- Once the White House signals monitoring intent through the EO, agencies like NIST and DHS become de facto standards-setters, pulling procurement and vendor behavior toward whatever those bodies define.
Third-order effects
- If the sequence holds — blueprint, then executive order, then OMB's annual reporting requirement — US AI governance consolidates around administrative machinery inside agencies rather than legislation from Congress, making career officials and OMB memos the real regulatory surface.
- The revolving door visible here (Nelson, Parker, Buchanan all moving between the White House and outside commentary) suggests AI policy expertise itself becomes a market, with ex-officials mediating between labs and government.
The trend: White House AI policy is institutionalizing step by step — from OSTP's voluntary Bill of Rights toward binding agency-level reporting and standards — with each former official's retrospective marking a stage in that hardening.