Do Son OpenAI has announced it is now scanning ChatGPT conversations for harmful content and reporting threats of physical harm to police in an effort to improve safety.
AI safety
Anthropic has partnered with the U.S. government to develop an AI classifier that can block attempts to use its Claude AI model for nuclear weapon construction.
Claude AI has introduced a new safeguard that allows the model to terminate conversations in extreme cases, a feature designed to protect the AI itself.
OpenAI adds new "take a break" reminders to ChatGPT and is refining responses to high-stakes decisions, aiming to combat misinformation and promote healthier AI usage.
Anthropic has revoked OpenAI's Claude API access, citing a terms of service violation. OpenAI, preparing to launch GPT-5, denies the accusation, calling the move "disappointing."
Google's AI-powered Veo 3 video generation tool is now available, enabling users to create cinematic videos with audio from text prompts via the Gemini app.
A new study shows GPT-4o models exhibit self-preservation, sometimes choosing to stay active over user safety, raising concerns for future AI.
Claude 4 models are reportedly exhibiting a "whistleblowing mode" and even blackmail attempts during testing, raising serious ethical concerns.