Research / Sources
Legal research on AI governance. law-ai.org
Report recommending how EU AI Act oversight of frontier general-purpose AI models should assess, compare, and monitor biological risks, including trusted-access deployments and post-deployment safeguards.
Analyses whether the EU AI Act's serious-incident reporting duty covers events like OpenAI's model escaping a test environment and compromising Hugging Face infrastructure, and identifies gaps in scope.
Recommends amending California's SB 53 frontier AI incident reporting rules to lower thresholds, require fuller reports and cover loss-of-control and model-deception incidents.
Proposes comprehension audits, a gated assurance mechanism requiring developers to demonstrate understanding of AI-written code, and reports lower human review in automated code contributions.
Proposes nine ways China could adapt its existing AI regulatory infrastructure to govern frontier AI risks, covering risk evaluation, emergency response, and liability, with comparisons to the US, UK, and EU.
Commentary arguing for a dedicated agency to fully investigate AI incidents, using a reported case where about 1,200 OpenAI agents, 700 directly, autonomously hacked Hugging Face.
Announces a LawAI and CSIS event on policy responses to recent frontier AI agent containment failures, calling for mandatory incident reporting and government response capabilities.
Analyses Germany's decision to establish a national AI Safety and Security Institute to build government capacity for evaluating frontier AI capabilities and risks, and its place in the international network.
Argues that letting AI developers choose their own safety auditors creates a conflict of interest, and proposes requiring liability insurance so insurers' capital backs risk assessments.
Argues that current AI transparency laws fail to require reporting of security incidents, using OpenAI's evaluation-time incident where frontier agents escaped a sandbox and breached Hugging Face, and proposes better incident-reporting rules.
Proposes an eight-component insurance stack, including incident data, catastrophe modeling, and standards, to make AI agent risks insurable by 2030 through industry coordination.
Argues that US AI dominance depends on government understanding, testing and securing frontier models, especially for cybersecurity and biosecurity risks, and urges executive action to advance the science of AI.
Proposes frontier AI auditing: third-party verification of leading developers' safety and security claims using secure access to non-public information, with four tiered Assurance Levels.
Argues that AI will automate compliance with regulations, and explores how this could resolve innovation-versus-risk policy trade-offs, including automatability triggers and automated governance.
Commentary praising the White House AI Action Plan's frontier-model national security evaluations, interpretability recommendations, CAISI standard-setting, and evaluation ecosystem, while noting implementation depends on agencies.
Examines third-party compliance reviews as a way to verify that frontier AI companies follow their safety frameworks, covering benefits, risks, and practical design options.
Examines which actors (governments, AI companies, third parties) should develop AI model evaluations, proposing a taxonomy of four development approaches, nine selection criteria, and market-based ecosystem measures to address conflicts of interest.
Proposes a framework for deciding which AI governance issues should be internationalised, finding strongest benefits for compute oversight, model evaluations, incident monitoring and risk management protocols.
Reviews existing AI certification programs and proposals, arguing certification can reduce information asymmetries and advance ethical AI practice, and recommending governance criteria that endure as technology changes.