# FAIR Labs > FAIR Labs (Fair Artificial Intelligence Research Labs) is a nonpartisan 501(c)(3) research institute in Washington, DC. It publishes independent, sourced research reports on AI safety, AI policy and governance, AI infrastructure (data centers, compute), child safety, algorithmic fairness, and AI in health. Reports may be quoted and reproduced with attribution for non-commercial purposes. Site: https://fairlabs.ai All reports: https://fairlabs.ai/reports.html Research summaries (HTML, one page per report): https://fairlabs.ai/research/ Contact: https://fairlabs.ai/index.html#contact ## Frontier Series - [The Backlash Against Data Centers](https://fairlabs.ai/research/frontier-special-the-backlash-against-data-centers) (FL-2026-SR2, September 2026): This Special Report of the Frontier Series audits the revolt against data center construction in the United States. It asks four questions the debate keeps answering by assertion: whether foreign money or messaging is behind the opposition; what the China angle actually threatens; whether the three claims most often made against data centers — that they are dumped on poor communities, that they drain the nation's water, and that they raise everyone's electricity bill — survive contact with the evidence; and what happens to communities, and to the country, when the answer to those claims is a moratorium rather than a rule. Its central finding: the grievances are mostly homegrown, the myths are mostly wrong, the genuine problems are mostly local and fixable by tariff and siting design, and the blunt instrument now spreading through statehouses would forfeit real fiscal dividends while relocating, not preventing, the facilities it targets. PDF: https://fairlabs.ai/reports/frontier-special-the-backlash-against-data-centers.pdf - [The Open Frontier](https://fairlabs.ai/research/frontier-special-the-open-frontier) (FL-2026-SR1, July 2026): On 16 July 2026 a Chinese lab, Moonshot AI, released Kimi K3 — a 2.8-trillion-parameter open-weights model that, on several public tests, trades blows with the leading American systems. This Special Report treats the release as the second “Sputnik moment” in eighteen months. It weighs the windfall for buyers of intelligence — selfhosting, fine-tuning, and a collapse in unit cost — against the harder question it poses to an American economy whose growth and market value increasingly rest on the premise that frontier capability stays scarce and dear. PDF: https://fairlabs.ai/reports/frontier-special-the-open-frontier.pdf - [The Safety Horizon](https://fairlabs.ai/research/frontier-vol01-the-safety-horizon) (FL-2026-01, July 2026): The flagship volume of the Frontier Series takes stock of artificial intelligence safety at mid-decade. Its central finding is a widening capability–assurance gap: the pace at which frontier systems gain autonomy and reach has outrun the pace at which we can evaluate, interpret, and control them. The report inventories what has gone right, maps where the gaps are widening, sketches four trajectories to 2030, and closes with an eight-point agenda that assigns concrete work to every stakeholder with a hand on the wheel. PDF: https://fairlabs.ai/reports/frontier-vol01-the-safety-horizon.pdf - [The Abundance Dividend](https://fairlabs.ai/research/frontier-vol02-the-abundance-dividend) (FL-2026-02, July 2026): The second volume of the Frontier Series prices the upside. Safe, broadly deployed AI could compress the cost of cognitive work the way electrification compressed the cost of power — but abundance is a dividend, not a default. It must be claimed through deployment, diffusion, and trust, and it is paid in proportion to all three. The report maps five abundance frontiers, diagnoses the diffusion bottleneck, argues that assurance is an economic input, and closes with a seven-point compact for collecting the dividend broadly. PDF: https://fairlabs.ai/reports/frontier-vol02-the-abundance-dividend.pdf - [Atlas of AI Risk](https://fairlabs.ai/research/frontier-vol03-atlas-of-ai-risk) (FL-2026-03, July 2026): Volume 03 is the series' field guide to what can actually go wrong. It maps the full risk surface of advanced AI as four continents — Misuse, Malfunction, Misalignment, and Structural drift — divided into twelve territories, each scored for severity and tractability. It traces the trade winds by which risks compound and convert across continents, and closes with watchtowers: observable early-warning indicators, a recommendations agenda, and a working glossary. The atlas's premise is simple: confusion about the map is itself a risk. PDF: https://fairlabs.ai/reports/frontier-vol03-atlas-of-ai-risk.pdf - [Governing the Frontier](https://fairlabs.ai/research/frontier-vol04-governing-the-frontier) (FL-2026-04, November 2025): The fourth volume of the Frontier Series is the governance handbook: it takes the risk surface Volume 03 maps and asks what a regulator actually signs, funds, or staffs to close it. It proposes a reporting bar for serious AI incidents, a verification profession to check vendor claims, a threshold ladder that scales obligation to capability, and a way through the coordination trilemma between speed, sovereignty, and consistency. The premise throughout: principles are step one, and the job is building instruments that work when no one is watching. PDF: https://fairlabs.ai/reports/frontier-vol04-governing-the-frontier.pdf - [The Alignment Agenda](https://fairlabs.ai/research/frontier-vol05-the-alignment-agenda) (FL-2026-05, July 2026): Alignment is often described as a single unsolved problem. This volume treats it as a portfolio of four bets — evaluation science, interpretability, scalable oversight, and control — that only pay reliably when held together, as defense in depth. For each bet the report assesses what exists, what it cannot yet do, and what maturity would look like, then proposes a funding and institutional program to make assurance compound industrially. It closes with twelve open problems that would move the field if solved this decade. PDF: https://fairlabs.ai/reports/frontier-vol05-the-alignment-agenda.pdf - [When Systems Can't Fail](https://fairlabs.ai/research/frontier-vol06-when-systems-cant-fail) (FL-2026-06, January 2026): The sixth volume of the Frontier Series is the deployment-safety field manual for anyone putting AI to work in systems that cannot simply be switched off: hospitals, banks, grids, courts, and the agencies people depend on daily. Its central argument is that the marginal risk of the next four years lives downstream, in ordinary deployments quietly standardizing on the same handful of models — and that the enemy is correlation, not any single flaw. The report names the fragility mechanisms, builds a resilience doctrine around graceful degradation and meaningful human authority, and specifies the assurance case every consequential deployment should be able to produce before going live. PDF: https://fairlabs.ai/reports/frontier-vol06-when-systems-cant-fail.pdf - [The Compute Compact](https://fairlabs.ai/research/frontier-vol07-the-compute-compact) (FL-2026-07, February 2026): The seventh volume of the Frontier Series treats compute — chips, clusters, energy, interconnect — as AI's most governable input and its most concentrated one. The report builds the case for compute-indexed thresholds as the backbone of Volume 04's governance ladder, confronts the security logic and limits of export controls, and makes the affirmative case for a publicly funded compute commons so that oversight capacity does not live only inside the firms being overseen. It closes with a decoded map of the compute stack for readers who have never had to think about a wafer fab before. PDF: https://fairlabs.ai/reports/frontier-vol07-the-compute-compact.pdf - [Work in the Age of Cognition](https://fairlabs.ai/research/frontier-vol08-work-in-the-age-of-cognition) (FL-2026-08, July 2026): The eighth volume of the Frontier Series examines what advanced AI actually does to work: not the disappearance of jobs but their recomposition, task by task, at a speed no previous technology transition has matched. Its central claim is that the transition rate — not the destination — is the policy variable that separates a humane adjustment from a brutal one. The report maps exposure and complementarity across task families, designs the mobility infrastructure a fast transition requires, and argues for worker voice and a fair share of the dividend. PDF: https://fairlabs.ai/reports/frontier-vol08-work-in-the-age-of-cognition.pdf - [The Trust Infrastructure](https://fairlabs.ai/research/frontier-vol09-the-trust-infrastructure) (FL-2026-09, July 2026): The ninth volume of the Frontier Series confronts the moment when convincing text, images, audio, and video can be fabricated at near-zero cost. Its central claim: societies need trust the way cities need water — as engineered infrastructure, not ambient good fortune. The report anatomizes the twin failure modes of gullibility and reflexive disbelief, designs a four-layer trust stack from provenance rails to a literate public, assesses what authentication can and cannot do, and assigns duties to the institutions whose vouching still holds. PDF: https://fairlabs.ai/reports/frontier-vol09-the-trust-infrastructure.pdf - [The Director's Dilemma](https://fairlabs.ai/research/frontier-vol10-the-directors-dilemma) (FL-2026-10, May 2026): The tenth and closing volume of the Frontier Series is written for the people who will be asked, after the fact, what the board knew and when. It argues that AI oversight is now squarely a board duty for every company — not just the ones building frontier models, since even the most AI-distant firm is at minimum dependent on vendors and infrastructure this series treats as a monoculture risk. The report names three postures a company can hold, builds the due-diligence playbook and committee architecture each requires, and closes the series with the self-assessment a board can run before its next AI-related decision. PDF: https://fairlabs.ai/reports/frontier-vol10-the-directors-dilemma.pdf ## AI Safety - [AI Safety Framework for Public Institutions](https://fairlabs.ai/research/ai-safety-framework) (2026): A structured approach to evaluating, procuring, deploying, and overseeing artificial intelligence in government and enterprise environments, with risk assessment matrices and implementation guidance. PDF: https://fairlabs.ai/reports/ai-safety-framework.pdf - [Evaluating Frontier Models: A Practical Guide](https://fairlabs.ai/research/evaluating-frontier-models) (2026): How institutions and technical teams should test large, general-purpose AI models — rigorously, adversarially, and continuously — before and after they are deployed. PDF: https://fairlabs.ai/reports/evaluating-frontier-models.pdf ## Child Safety - [Children in the Age of Generative AI](https://fairlabs.ai/research/children-generative-ai) (2026): Emerging threats to children on AI-powered platforms, and the safety-by-design measures that platforms, schools, and policymakers must put in their path. PDF: https://fairlabs.ai/reports/children-generative-ai.pdf - [AI Companions and the Developing Mind](https://fairlabs.ai/research/ai-companions-children) (2026): What the evidence says about children forming relationships with conversational AI, and the design standards that would make companion systems safe. PDF: https://fairlabs.ai/reports/ai-companions-children.pdf ## Fairness & Bias - [Auditing Algorithmic Bias in Practice](https://fairlabs.ai/research/bias-audit-toolkit) (2026): A practical toolkit for identifying and mitigating bias in deployed AI systems — scoping, measurement, testing, remediation, and governance. PDF: https://fairlabs.ai/reports/bias-audit-toolkit.pdf - [AI Transparency Standards](https://fairlabs.ai/research/ai-transparency-standards) (2025): Normative standards for the transparency and explainability of AI systems, with technical specifications, documentation schemas, and a tiered conformance framework. PDF: https://fairlabs.ai/reports/ai-transparency-standards.pdf ## AI Policy - [AI Policy Recommendations for Congress](https://fairlabs.ai/research/congressional-ai-policy) (2025): Detailed, nonpartisan recommendations for federal AI regulation, governance, and oversight — built to endure across administrations and generations of technology. PDF: https://fairlabs.ai/reports/congressional-ai-policy.pdf - [Ethical AI Development Guidelines](https://fairlabs.ai/research/ethical-ai-guidelines) (2025): A practical program for building AI responsibly — translating ethical principles into disciplined process, ethical review, and genuine stakeholder engagement. PDF: https://fairlabs.ai/reports/ethical-ai-guidelines.pdf ## AI & Health - [Mixed Reality in Medical Training: Evidence Review](https://fairlabs.ai/research/mixed-reality-medical-training) (2025): Reviewing the evidence on volumetric and mixed-reality communication for remote medical procedure training — what it can show, what it has shown, and what remains unproven. PDF: https://fairlabs.ai/reports/mixed-reality-medical-training.pdf - [Responsible AI in Telehealth](https://fairlabs.ai/research/responsible-ai-telehealth) (2025): Guidance for health systems adopting AI-assisted telehealth — safety, oversight, consent, and equity, from procurement to the point of care. PDF: https://fairlabs.ai/reports/responsible-ai-telehealth.pdf ## Embodied AI - [Embodied Intelligence: Safety Beyond the Screen](https://fairlabs.ai/research/embodied-intelligence-safety) (2025): When AI leaves the browser and enters physical and virtual space — robotics, autonomous systems, extended reality, virtual humans — a new class of safety questions follows it. PDF: https://fairlabs.ai/reports/embodied-intelligence-safety.pdf - [Virtual Humans and the Ethics of Presence](https://fairlabs.ai/research/virtual-humans-ethics) (2025): Photorealistic volumetric avatars and AI-driven virtual humans raise questions of consent, identity, and deception — toward a framework for honest embodied representation. PDF: https://fairlabs.ai/reports/virtual-humans-ethics.pdf ## Other - [Dev Rankings](https://fairlabs.ai/devrank/): FAIR Labs' ranking of developer tools and packages as selected by AI coding agents. - [Blog](https://fairlabs.ai/blog/): research notes.