Policy Ethics

OpenAI is advocating for a 'reverse federalism' approach to AI regulation, where state-level laws serve as laboratories to develop a comprehensive national safety framework. This bottom-up strategy aims to foster innovation while ensuring responsible AI development. Will this model work?

Anthropic apologized for using covert guardrails in Claude Fable 5 that altered responses to block model distillation without notifying users.

Nvidia, Microsoft, and Meta have urged Washington not to impose strict regulation on open-weight AI models, arguing such rules would favor large closed-model providers, according to CNBC.

AI startup founders are urging the Trump administration not to ban Chinese open-weight models, warning it could create a duopoly among large closed-source AI firms, per Politico.

OpenAI's new report reveals that up to 80% of European occupations could be impacted by AI. The study details which jobs face automation and where new growth will emerge.

HCCF has proposed a .self top-level domain letting users self-host their data and identity outside corporate platforms, though the plan is still an early proposal, not an approved TLD.

Anthropic mandates a non-optional 30-day data retention policy for Claude Mythos 5 and Fable 5 models, ending zero data retention for enterprise users.

OpenAI and the Walton Family Foundation are launching "AI Skills Jams," hands-on workshops to train K-12 educators in practical classroom AI use.

OpenAI has published a framework governing its work with government and national security clients, ruling out autonomous weapons and requiring human oversight.

Digital rights group Fight Chat Control says the EU's 'Chat Control 2.0' proposal would require AI scanning of private messages before encryption, raising fears of mass surveillance.

Developer Susam Pal has proposed three 'Inverse Laws of Robotics' to guide human behavior when interacting with AI systems and prevent uncritical reliance.

OpenAI will let a U.S. government committee approve or deny high-level access requests for its upcoming GPT-5.6 model, according to The Washington Post.