AWS outlines a phased responsible AI roadmap for agentic systems, extending governance beyond model training and inference to tool use, multi-step reasoning, and autonomous actions. It ships an open RAI assessment skill on GitHub so platform and application teams can audit their governance posture with their own agents.
Responsible AI
负责任 AI
Governance practices covering fairness, explainability, safety, and compliance for AI systems.
Curated Intelligence
Anthropic has opened a local office in Singapore, partnering with GovTech, Grab, and DBS Bank. The collaboration focuses on assessing AI safety and large model risks to promote responsible deployment across public and private sectors. Data indicates Singapore ranks second globally in per-capita Claude usage, significantly exceeding population-based predictions.
OpenAI has officially joined the EU Code of Practice on AI, committing to responsible development. The move aims to foster innovation, infrastructure, and economic growth through partnerships with European governments.
Hugging Face released updated community content guidelines to address specific abuse risks associated with machine learning models. The policy refines moderation standards for generative AI content, emphasizing a balance between open collaboration and responsible development.
OpenRAIL aims to establish open and responsible licensing standards for AI models, balancing the open-source ecosystem with security risks. The framework seeks to regulate usage boundaries through legal clauses to prevent misuse and promote compliant development. It is currently in the conceptual and early exploration phase, with specific implementation details yet to be finalized.
OpenAI released a policy research paper outlining four strategies to foster industry cooperation on AI safety: communicating risks, technical collaboration, increased transparency, and incentivizing standards. The report aims to address collective action problems caused by competitive pressures that may lead to under-investment in safety, ensuring AI systems remain safe and beneficial.