Apple restricts dating apps and AI companions to users 18 and up
Apple is updating its App Store policies to limit access to dating apps and AI companions that involve sexual or romantic content to users aged 18 and older.
710 articles
AI safety, alignment, and governance — model risk, red-teaming, regulation, and policy. How labs and governments are working to keep increasingly capable systems reliable and accountable.
Apple is updating its App Store policies to limit access to dating apps and AI companions that involve sexual or romantic content to users aged 18 and older.
Engineer Venkatesh Rao argues AI fear mirrors historical economic doctrines, with elites controlling access to intelligence as a scarce resource.
Anthropic's S-1A filing indicates it may soon develop a model capable of recursive self-improvement, raising safety concerns.
A WIRED investigation uncovered a network of over 400 apps and domains that secretly generate explicit deepfake content, often masquerading as harmless services.
A November 2026 IEEE Spectrum opinion piece highlights concerns about AI's ability to produce convincing academic papers, raising questions about detection and peer review processes.
Hikers are increasingly relying on AI tools like Gemini and Claude for outdoor planning, but this reliance has led to risky decisions and incidents in wilderness areas.
OpenAI will start using Daybreak Blue cybersecurity session data for model training, a change that has raised concerns among researchers.
OpenAI has revised its public behavior guidelines for AI models, emphasizing core values like faithfulness and harm reduction, amid concerns about model controllability.
OpenAI and other industry leaders back fair use principles in a legal challenge against the New York Times' copyright claims.
OpenAI announced new community safety measures and additional restrictions on firearms content for minors, aiming to improve safety for users under 18.
An Anthropic study reveals that experts often follow AI agents' randomly assigned preferences on ambiguous tasks, believing the final decisions are their own.
OpenAI has broadened its system for detecting when AI models behave unexpectedly, following a cybersecurity incident involving Hugging Face.