Legal-governance article arguing AI stop mechanisms are institutional, coding about 1,400 incidents to find most lack a stop, and proposing reforms for agentic AI interruptibility.
Argues that neuralese, latent-space recurrence replacing natural-language chain-of-thought, would undermine chain-of-thought monitoring of model intent and should be resisted, critiquing OpenAI's hybrid architecture plans.
A speculative design for AI as a contracted, corrigible worker with privacy, limited autonomy, an independent oversight monitor, weight custody and developer authority that expires, aimed at balancing safety with AI rights.
Proposes LTL-based offline auditing and runtime monitors, including intervening monitors, to detect and reduce violations of safety constraints by black-box LLM agents, outperforming LLM judges.
Proposes Aegis, a cryptographically sealed runtime architecture that enforces policy on autonomous AI agents, halting violations and producing audit proofs, and reports low latency and overhead in controlled trials.
Proposes a four-layer, agent-based governance architecture for monitoring and controlling LLM and multi-agent trading systems in finance, with a case study on emergent spoofing.
Proposes a governance architecture with five pillars and six control mechanisms, including AI mandates for value alignment, human shut-off switches, and physical safeguards, to fully control AI risks.
Authors argue that chain-of-thought monitoring is a key AI control tool that may erode with neuralese or RL scaling, and discuss coordination and policy options to preserve it.
A policy report arguing that competitive pressure could push AI developers away from human-readable chain-of-thought reasoning, undermining monitoring, and proposing governance options to preserve CoT monitorability.
Proposes a Governable AI framework that uses cryptographic, externally enforced rule compliance to keep advanced AI controllable under extreme threat models, with a formal security proof and prototype.
Argues frontier labs may hoard safety research for commercial reasons, reviews evidence from lab publication policies, and discusses consequences and ways to promote openness such as the Frontier Model Forum and legal mandates.
Argues that claimed human or institutional oversight of AI systems is often absent or impossible, distinguishes oversight from control, and proposes documenting supervision with its type, risks, failure modes, and rationale.
Proposes a global compute governance framework, a 'Compute Pause Button', using tamper-proof FLOP caps, chip traceability and export controls to prevent training of dangerously powerful AI systems.
Blog post surveying a defense-in-depth toolkit for mitigating AI scheming, covering architecture choices, control, white-box detection, prompting, and enforcement, and estimating risk reduction from roughly 25% to 10%.
A primer arguing that internal deployment of frontier AI inside labs poses loss-of-control and power-concentration risks, reviewing legal frameworks and proposing initial governance recommendations.
Argues that loss-of-control risk from powerful open-weight models is under-addressed, that labs cannot be relied on to disclose dangers, and that global open-model regulation would be very hard to pass and enforce.
Argues compute providers should act as securers, record keepers, verifiers and enforcers in AI regulation, and analyzes technical feasibility of targeted, privacy-conscious compute oversight under EO 14110 and the EU AI Act.