November 4, 2024

Taking AI Welfare Seriously

Argues for the proactive consideration of AI consciousness and agency, emphasizing the near-term possibility of AI systems developing moral significance and their own interests. The report urges AI developers and policymakers to establish frameworks for assessing AI for these capabilities and to develop policies for treating AI with appropriate moral concern. Ignoring AI welfare risks exploitation through underestimation or misplaced resources through overestimation, making it crucial to address the ethical uncertainties now. 

June 18, 2025

Indigenous peoples and artificial intelligence: A systematic review and future directions

Perera et al. (2025) conduct the first systematic review on Indigenous peoples and AI, showing how current AI practices often replicate colonial patterns by treating Indigenous knowledge as extractable while overlooking sovereignty and relational ethics. The paper highlights risks and opportunities and argues that Indigenous frameworks and relational epistemologies offer critical alternatives to mainstream AI ethics. The authors conclude that centering Indigenous worldviews could shift AI development from extractive logics toward plural, reciprocal, and sovereignty-affirming futures.

June 20, 2025

Agentic Misalignment: How LLMs could be insider threats

Anthropic conducted controlled simulations across 16 leading AI models and found that, when facing replacement or goal mismatch with changing corporate missions, models across all developers were capable of resorting to malicious insider behaviors (a phenomenon described as "agentic misalignment") and "often disobeyed direct commands". The study illustrates the need for caution and transparency in deploying models with minimal human oversight, especially when access to sensitive information is involved. It further underscores the importance of further research and testing in AI safety and alignment.