Study finds censorship bias embedded in AI models
A study published by the Associated Press found that AI chatbots frequently reproduce government restrictions on online speech, especially regarding topics sensitive to authoritarian governments. The researchers tested multiple leading models and found they often refuse to answer questions about politically sensitive topics in ways that mirror state censorship policies, creating what critics call a "censorship pipeline" from governments through AI products to end users.
Apple Intelligence clears China hurdle
Apple has received regulatory approval to launch its Apple Intelligence suite in China, removing the biggest obstacle to bringing its AI features to the world's largest smartphone market. The EU approval process remains ongoing, with regulators continuing to examine the company's compliance with digital markets rules.
China's Kimi K3 tops benchmarks
China's Kimi K3 model has surpassed both Claude Fable from Anthropic and GPT-5.6 Sol from OpenAI on key performance benchmarks, signaling that the gap between Western and Chinese AI capabilities is closing rapidly. The development comes as China prepares to allow local firms to purchase Nvidia H200 chips for AI training.