Study Finds Chinese Models Echo Official Views on Taboo Topics
An Aleph Alpha benchmark finds Qwen, DeepSeek and Kimi often refuse or reproduce official positions across 967 politically sensitive prompts.
All briefings and quick hits published that day.
An Aleph Alpha benchmark finds Qwen, DeepSeek and Kimi often refuse or reproduce official positions across 967 politically sensitive prompts.
Aleph Alpha releases a German-English MoE model with 78.1 billion parameters, 3.46 billion active per token and Apache 2.0 licensing.
David Robinson, who led OpenAI’s Preparedness Framework and supervised safety reports for 12 frontier launches, says the company’s deployment culture has outpaced the safeguards needed for increasingly capable AI systems.