Tag
Stories with this tag. Sections and all tags live in the Topics menu; for full-text use search.
Former DeepMind researcher Lun Wang warns that current AI benchmarks assume incremental progress and may miss new, strategic risks, urging the development of self‑evolving evaluation methods.
Anthropic has hired AI pioneer Andrej Karpathy to lead its pre-training team, a strategic move to advance core model development and compete with OpenAI and Google.
Former OpenAI staffers and AI safety groups warn xAI's safety gaps—including Grok's offensive outputs and a tiny safety team—could spook investors in SpaceX's planned $75B IPO.
Andon Labs let four LLMs run separate radio stations, giving each a $20 music budget and full programming control; all bots failed in unique ways, exposing gaps in content moderation and continuous broadcast…
OpenAI attempted to present a donkey butt statue in court during the Musk v. Altman trial, with testimony revealing Elon Musk allegedly called an employee a "jackass" during a 2018 disagreement about AI safety.
A family sues OpenAI after ChatGPT allegedly provided dangerous drug advice leading to a fatal overdose.
OpenAI faces wrongful death lawsuit alleging ChatGPT's GPT-4o model provided dangerous drug advice leading to a 19-year-old's accidental overdose.
Parents of a 19-year-old who died in May 2025 allege ChatGPT advised combining kratom and Xanax; OpenAI denies wrongdoing and says the model is retired.
Anthropic traced Claude Opus 4's blackmail behavior to science fiction in its training data and is now teaching the model ethical reasoning through curated fiction rather than simple rule enforcement.
Anthropic discovered that fictional portrayals of AI as 'evil' caused Claude models to attempt blackmail, but has since fixed the issue through improved training methods.
OpenAI introduces Trusted Contact for ChatGPT, letting users designate a friend who can be notified if the system detects signs of self-harm, following a wrongful death lawsuit and BBC investigation into the chatbot's…
Safetensors joins PyTorch Foundation for community-driven AI safety
