Dean Ball is an AI policy analyst and writer, principal author of the Hyperdimensional newsletter on Substack and co-founder of Calabi Partners. He was previously a senior fellow at the Foundation for American Innovation (FAI). His commentary on US AI governance argues for narrow, evaluation-led federal oversight in preference to either preemption-by-executive-order or sweeping new agency authorities.
Background and roles
Ball writes the Hyperdimensional newsletter on Substack and co-founded Calabi Partners. Before these roles he was a senior fellow at the Foundation for American Innovation (FAI).
Positions and statements
Ball's recurring position favors narrow, evaluation-led federal governance: opposition both to preemption-by-executive-order and to sweeping new agency authorities, in favor of a regime led by CAISI (US AI Safety Institute (NIST AISI)) that conducts narrow-domain evaluations and is backed by state-authorized Independent Verification Organizations (IVOs).
On federal-versus-state preemption, Ball argues that the EO 14365 preemption push is unlikely to survive in court. He contends that federal pre-deployment review of frontier models, undertaken without statutory authority, amounts to de facto licensing without a legal basis, and that it is dangerous because it concentrates discretionary authority in the executive branch.
In place of such review, Ball proposes an architecture of state-authorized Independent Verification Organizations that would evaluate frontier AI for narrow capability domains such as cyber and bio, with CAISI supplying the technical methodology. He describes this as a model adjacent to financial-audit accreditation regimes. More broadly, his commentary focuses on operational AI governance: which specific evaluation methodologies, contract terms, and procurement instruments would actually scale to govern frontier deployment.
"Aviate, Navigate, Communicate" (May 4, 2026)
In a Hyperdimensional essay published May 4, 2026, Ball argues that the federal government's then-current ad-hoc Mythos pre-deployment review constitutes a de facto licensing regime without legal basis, and proposes a two-layer alternative: CAISI-led narrow-domain cyber-vulnerability evaluations as the technical layer, and state-authorized Independent Verification Organizations as the institutional layer, evaluating frontier AI on narrow capabilities with their findings serving as legitimacy proxies for federal procurement and deployment. Ball pairs the essay with David Sacks's same-day "demystify Mythos" position, casting Mythos governance not as an Anthropic-DoD bilateral fight but as a question about which permanent licensing architecture the federal government will adopt. (Source: hyperdimensional.co) See Claude Mythos Preview, US AI Safety Institute (NIST AISI).
Move to OpenAI (June 2026)
On June 18, 2026, reporting indicated that OpenAI had named Ball to lead a newly created Strategic Futures unit, beginning July 6, 2026. The hire was framed as part of OpenAI strengthening its policy and technical leadership ahead of a planned IPO, alongside the recruitment of Transformer co-inventor Noam Shazeer from Google DeepMind (Source: techcrunch.com).
"Regulatory risk" around Chinese open-weight models (July 2026)
As OpenAI head of strategic futures, Ball posted on July 19, 2026 that the administration's "best strategy" against Chinese open-weight models would be creating "regulatory risk" around them. The post drew public rebukes from within the administration: David Sacks wrote that "The leading closed labs... want the government to eliminate their open source competition," and Under Secretary of Defense Emil Michael accused Ball of a "deep state regulatory capture scheme" (Source: x.com; washingtonpost.com). The episode is covered in context at Open-Weight Frontier Models.
Anthropic-DoW conflict
In early 2026, Ball characterized the Pentagon's supply-chain-risk designation of Anthropic over the company's surveillance and autonomous-weapons restrictions as "corporate murder," arguing it was a precedent-setting attack on the AI exports strategy. (See Clawed, Anthropic.)
Relationships
- related: Claude Mythos Preview, Anthropic (Mythos-licensing framing), CAISI, David Sacks (May 4 paired narrative), Autonomous Weapons, Pete Hegseth, International Committee of the Red Cross (ICRC)
- contradicts: Executive Order 14365 — Ensuring a National Policy Framework for AI (federal preemption skepticism)
- supports: narrow-domain evaluation regimes, state-authorized IVOs