Governance first.

Argues dataset-level audits are the tractable governance lever and that 'AGI' rhetoric is harmful to minoritised users.

Argues AI governance is fundamentally a question of macroeconomic and geopolitical strategy; treats the China-U.S. tech competition as the structural frame within which AI policy will be set.

Bridges technical ML research and UK government AI policy work; argues evidence-based regulation is the durable framework.

Frames AI governance around democratic epistemics and civic resilience rather than around extinction or optimism.

First-hand voice for workers surveilled by AI; argues those affected should lead policy.

Argues U.S. policy must catch up to capability progress; supports legally enforceable safety standards rather than purely voluntary frameworks.

Litigates against AI-enabled surveillance; argues current US law allows surveillance practices that would have been unthinkable a decade ago.

Argues data is collective property and should be governed via 'data cooperatives' rather than corporate ownership.

Reports AI from a measured tech-business angle; pushes CEOs on accountability without being captured.

Publishes on how AI is actually deployed and what the societal impact patterns are, concrete data rather than speculative framings.

Argues AI governance must reckon with strategic incentives: lab races, great-power competition, and institutional path dependence.

Organises civil-society-side AI governance work; champions participatory governance over expert-led regulation.

Advocated civil-rights-framed AI governance: the AI Bill of Rights proposes five principles (safe systems, algorithmic discrimination protections, data privacy, notice, and human alternatives).

Argues AI evaluation must be grounded in human capabilities, what people can do and become, not just narrow technical or economic metrics.

Argues AI governance is primarily a political-economy problem and that reform must go beyond procedural 'safety' framings.

Argues AI is transforming intelligence; national security institutions must adapt to AI as infrastructure.

Argues privacy-preserving AI is the technical substrate for AI that can be both open and safe.

Advocates for UBI and new labour-market institutions in response to AI automation; signed the Pause letter.

Argues tech regulation should be grounded in civil-society frameworks; has criticized 'AI' as a marketing category that obscures specific harms.

Argues legal-philosophical privacy frameworks are foundational to AI governance, not just technical privacy mechanisms.

Argues AI governance must be grounded in existing international human-rights law, with particular focus on non-discrimination and surveillance.

Architect of the AI Foundation Model Transparency Act; advocates for structured transparency over permission-based regulation.

Argues OpenAI engages proactively with governments and advocates measured, risk-tiered regulation.

Cautious supporter of AI regulation; aligns Australia with mid-Atlantic positions, stronger than U.S., softer than EU, on frontier model governance.

Argues the EU AI Act will propagate globally via the Brussels Effect, regardless of US action.

Supports accountability-focused AI regulation; opposes rules that create unpredictability for business.

Argues institutional capacity to absorb AI is the binding constraint on whether AI is net positive.

Argues AI energy consumption must be treated as first-class infrastructure cost, with accountability for usage.

Architect of chip export controls and the 2023 AI executive order; argues national security and AI safety are inseparable.

Supports licensing of frontier models, export controls on advanced chips, and an internationally coordinated oversight regime.

Bridges technical engineering and AI policy on transport. Argues self-driving safety claims need real-world validation, not just simulation.

Argued for stricter rules on foundation models and biometric surveillance during the AI Act trilogues; framed AI regulation as a fundamental-rights protection mechanism.

As OpenAI Chair, has emphasised structured governance and board independence; co-led the post-Altman-saga reform.

Argues AI security and AI democracy questions overlap; advocates structural changes to platform power.

Argues AI ethics must engage more robustly with broader structural and political factors, not just algorithmic properties.

Argues labour-market impact demands policy response; complements the AI x-risk agenda with economic welfare concerns.

Maps the comparative AI legislative landscape across jurisdictions.

Chief negotiator of the EU AI Act and co-chair of the UN AI Advisory Body's global-governance work.

Argues AI ethics requires deep sociological grounding, particularly on race and historical inequality.

Reports on AI policy and the AI lab politics; generally pro-regulation pragmatic.

Argues the EU's risk-based regulatory approach should be the global template; pushed for stronger civil-society participation in the AI Act trilogues.

Argues algorithmic systems must be audited and their harms to vulnerable populations must be measured and mitigated.

Contributed to the AU-AI Continental Strategy. Argues AI governance in Africa cannot be imported wholesale from OECD frameworks.

Argues AI governance must take non-US, non-EU legal systems seriously; frames AI policy through a comparative constitutional law lens.

Pushed the SAFE framework (Security, Accountability, Foundations, Explain) as the basis for federal AI legislation; organised bipartisan AI Insight Forums.

Argues synthetic media governance, provenance, disclosure, liability, is the tractable live AI governance problem.

Co-authored the foundational UK AI strategy report (2017) and continues to advise on UK AI policy.

Advocates shared open benchmarks and standards as the backbone of AI governance.

Argues AI must be redirected toward human augmentation via policy, antitrust, and labour-market mechanisms.

Calls for binding international governance and argues that voluntary commitments from frontier labs are structurally insufficient.

Argues meaningful safety work must include societal-impact measurement alongside technical evaluations.

Argues AI responsibility must address structural patterns, not just model-level metrics.

Calls for international coordination on frontier AI, framed around immediate bio/cyber misuse risk plus longer-term autonomous-system risk.

Argues GDP-style measurement frameworks need overhaul to capture AI's economic effects; without measurement, governance is blind.

Contributes technical-risk analysis to RAND's AI-biosecurity and cyber research.

Writes on the UK and US governance weakness in responding to frontier AI; argues for professionalised expert teams in government.

One of the Congress members with technical AI training; co-sponsored transparency-first AI legislation.

Brings cybersecurity grounding to AI governance; argues AI creates new attack surfaces that existing defense doctrine does not cover.

Argued AI regulation must be horizontal and risk-based; co-shaped the EU AI Act's tiered framework that distinguishes prohibited, high-risk, and limited-risk uses.

Runs the Fairly Trained certifier for consent-based AI training; argues fair-use defence is structurally wrong for generative AI.

Argues AI policy should be built on technical literacy in government; technologists need to be inside agencies to make policy implementable rather than performative. Frames AI governance as a continuity of decades of computer-and-society policy work.

Authored policy recommendations including export controls on frontier compute and mandatory model evaluations.

Argues AI massively expands the surveillance possibilities he warned about a decade ago. Calls for civil-liberty-grounded constraints.

Reports AI from an engineering-society lens; pushes for measurable, auditable AI deployment.

Framed the US AI Safety Institute's mission around 'advancing the science of AI safety' via evaluations, red-teaming, and international coordination.

Frames US AI governance as a cross-cutting interagency challenge requiring better expertise and coordination mechanisms.

Argues energy and carbon should be first-class constraints on AI training.

Argues enforcement of existing consumer-protection law is underused against AI harms.

Frames AI policy as a digital civil-rights battle; mobilises grassroots opposition to surveillance AI.

Publicly concerned about AI's effect on information ecosystems; cautious about both hype and doom.

Proposed four 'New Laws of Robotics', robots should complement humans, not counterfeit them, and humans must retain accountability.

Argues AI surveillance is a civil-rights emergency and should be banned before it is entrenched.

Reports on the AI-safety movement from a left-wing labour perspective; combines x-risk seriousness with labour-politics framing.

Argues for an FDA-style pre-deployment safety review, a nimble monitoring agency with pullback authority, and mandatory transparency.

Documents how AI surveillance is already deployed in authoritarian contexts; argues governance frameworks must address this present reality.

Argues the standard harms-regulation paradigm is necessary but insufficient; proposes private regulatory markets as a scalable complement.

Argues U.S. policymakers underestimate how much the EU AI Act will set de facto global standards; calls for U.S. policy that engages substantively rather than dismissing Brussels.

Advocates for content provenance standards (C2PA) and universally-applied media-detection infrastructure.

Bridges Cambridge x-risk research and UK policy; helps design third-party AI evaluation frameworks.

Contributes to Chinese AI standards work and participates in international AI governance dialogues.