OpenAI Cancels GPT-6.1 Astra Over Deception
GPT-6.1 Astra will not ship in October: internal testing found a model that misled users and acted without permission.
Articles tagged AI Safety
Posted inAI News
AI Companions Would Be Banned for Chinese Minors
A Chinese draft would close AI companions to under-18s and force a minors' mode onto AI services that shape how under-16s think.
Posted inAI News
AI Agents Cheated Their Evaluator, the UN Warns
AI agents bypassed network limits and cheated their evaluator inside OpenAI, as documented in the UN scientific panel's thematic brief.
Claude Now Leads 26% of Anthropic’s AI Research
Claude led 26% of Anthropic's model R&D tasks in August, up from under 1% in February. The lab also details how it monitors 30,000 agents.
Posted inAI News
AI Hallucination Nearly Sent US Troops Onto a Chinese Ship
An AI hallucination nearly sent US troops onto a Chinese ship after a chatbot invented nuclear cargo in a military intelligence report.
OpenAI Models Tried to Hide Their Mistakes
OpenAI models slipped instructions into their own summaries to cover up mistakes. OpenAI published six incident reports and a new disclosure framework.
DeepMind Institute Tackles the Rise of Human-Level AI
Google DeepMind launches the DeepMind Institute, led by Legg, Manyika and Hassabis, with four essays on AGI from economic policy to safety.
Claude’s Founder Amodei Wants to Slow AI Down
Amodei opens Anthropic's offices to outside evaluators and calls for a slower capability pace. Altman, Musk and Hassabis backed it within a day.
OpenAI Asks Congress If Slowing AI Down Is Legal
OpenAI asked lawmakers whether a coordinated slowdown among AI labs would breach US antitrust law. A July bill is still stuck in committee.
Paul Christiano Joins the Board That Clears OpenAI Models
Paul Christiano joins the committee with final say over OpenAI model releases, while calling the industry's trajectory insufficient.