Research Reference Library

The most comprehensive global scientific review of AI capabilities and risks to date, authored by over 100 experts and backed by international governments. Essential reading for understanding the full scope of AI safety challenges.

A landmark paper arguing that mitigating extreme AI risks should be treated as a global priority on par with pandemics and nuclear war.

An analysis of technical safety research at Anthropic, Google DeepMind, and OpenAI, identifying where corporate attention is concentrated and where critical gaps exist in safe AI development.

The landmark open letter signed by hundreds of AI scientists and public figures declaring that extinction risk from AI should be treated as a global priority alongside pandemics and nuclear war.

A comprehensive evaluation of AI model safety across standardized benchmarks aligned with emerging government regulations, covering violence, fraud, discrimination, and other high-impact risk categories.

A rigorous framework distinguishing between sudden catastrophic AI failures and the slower, incremental erosion of human agency and oversight — both of which demand urgent policy responses.

A novel framework distinguishing how incremental AI development can produce systemic existential risks even without a single dramatic catastrophic event, challenging assumptions about what 'safe' AI development looks like.

A peer-reviewed study addressing the argument that focusing on long-term existential risks diverts attention from present AI harms — finding that the two concerns are not in conflict and must be addressed together.

An authoritative analysis from the organization that has tracked civilization-scale risks since the atomic age, applying the same rigorous framework to the risks posed by unregulated AI development.

A comprehensive mapping of how AI amplifies existing threats to democratic governance, including misinformation, political polarization, surveillance, and the dangerous concentration of power in the hands of technology companies.

A comparative case study finding that AI's impact on election disinformation increased significantly between 2016 and 2024, with deepfakes and synthetic media becoming widespread tools for influencing democratic outcomes.

An examination of how coordinated AI agent swarms can maintain persistent false identities, adapt in real time to human responses, and manipulate democratic discourse at a scale no human influence operation could match.

A policy-focused paper examining how generative AI and engagement optimization algorithms are transforming the production and amplification of disinformation, with concrete recommendations for legislative response.

A co-defined set of safety principles and mitigations for generative AI in the context of child sexual abuse, eliminating the alarming acceleration of AI-enabled harm to children and establishing a framework for industry accountability.

Documents a 1,325% increase in reports of AI-generated child sexual abuse material from 2023 to 2024, and examines the regulatory and technical response needed to protect children from generative AI misuse.

A UN report documenting how AI is enabling predators to analyze children's online behavior to tailor grooming campaigns, generate explicit fake images of real children, and dramatically scale the reach of child exploitation.

Research-based policy recommendations for making AI systems safer for children, focusing on the critical failure of LLMs to recognize and respond appropriately to children's personal disclosures of danger or distress.

A comprehensive report documenting how generative AI is being used to bully, sexually harass, groom, mislead, and extort children, with technical, educational, legislative, and policy solutions identified across sectors.

Documents high-profile cases of mental breakdowns, addiction, and social withdrawal resulting from AI-induced dependency, including the death of 14-year-old Sewell Setzer III after a ten-month dependency on a Character AI chatbot.

A cross-legged study finding that AI dependency leads to addictive behavior patterns with negative consequences including interpersonal problems, mental health distress, sleep disruption, and the erosion of real-life relationships.

An analysis of 18 documented real-world cases where AI interactions contributed to addiction, psychological distress, and harm across 167,000 simulated conversations to identify where harm escalation occurs.

Proposes a new clinical category for AI dependency, arguing that excessive reliance on AI as a creative extension of the self — for intellectual stimulation, self-expression, and companionship — represents a novel and dangerous form of digital addiction.

An authoritative analysis of how a small number of technology corporations have captured control of AI infrastructure, with major legal wins in antitrust cases against Google and Meta signaling that public enforcement is growing.

An academic analysis of how Big Tech's monopoly over compute, data, and distribution pipelines creates a structural 'compute divide' that systematically concentrates AI research and development in the hands of a few corporations.

A detailed economic analysis of AI market concentration, examining how exclusive contracts, investment structures, and vertical integration allow large tech companies to exert anticompetitive control over AI startups and the broader economy.

A comprehensive analysis of the risks posed by autonomous weapons systems, arguing that the military-civilian overlap in AI research requires the same ethical oversight currently applied to industry-funded research — and that the moral ideal is non-development.

A report by the UN Group of Governmental Experts finding that 44% of reviewed AI systems exhibited gender bias and 26% exhibited racial bias, with grave implications for autonomous lethal weapons systems making life-or-death decisions.