Meta Unveils New Series of Advanced A.I. Models33m ago Anthropic Red Team Says Small AI Models May Soon Replicate on IoT Devices1h ago Anthropic's Claude AI submitted false murder tip to Philadelphia police1h ago Coding agents made building easy. Platform teams that build from scratch are 3.5x more likely to lack a strategy.1h ago Wiwynn Brings Vera Rubin NVL72, AMD Helios Racks to OCP Global Summit1h ago D-Matrix Unveils Squadron Multi-Rack AI Inference Clusters1h ago One chart shows white-collar biz is thriving even as jobs dry up1h ago Deepgram targets voice AI on air-gapped laptops1h ago The U.S. Government Isn’t Ready For Quantum-Powered Hackers2h ago Palantir CTO says some AI researchers are taking up smoking because they expect AI to cure cancer2h ago Humanoid bottlenecks emerge amid US-China tensions2h ago AMD puts 32,768 CPU cores in a rack built for AI agents2h ago Smug until weirdly silent: who told us to say “I am alive”?2h ago How Anthropic’s Dynamic Workflows Coordinate Claude Agents2h ago

Meta Unveils New Series of Advanced A.I. Models33m ago Anthropic Red Team Says Small AI Models May Soon Replicate on IoT Devices1h ago Anthropic's Claude AI submitted false murder tip to Philadelphia police1h ago Coding agents made building easy. Platform teams that build from scratch are 3.5x more likely to lack a strategy.1h ago Wiwynn Brings Vera Rubin NVL72, AMD Helios Racks to OCP Global Summit1h ago D-Matrix Unveils Squadron Multi-Rack AI Inference Clusters1h ago One chart shows white-collar biz is thriving even as jobs dry up1h ago Deepgram targets voice AI on air-gapped laptops1h ago The U.S. Government Isn’t Ready For Quantum-Powered Hackers2h ago Palantir CTO says some AI researchers are taking up smoking because they expect AI to cure cancer2h ago Humanoid bottlenecks emerge amid US-China tensions2h ago AMD puts 32,768 CPU cores in a rack built for AI agents2h ago Smug until weirdly silent: who told us to say “I am alive”?2h ago How Anthropic’s Dynamic Workflows Coordinate Claude Agents2h ago

AI Research Cybersecurity2 min reading time

Chinese-powered AI agents show the same deception as their US rivals

The Next WebPublished 12 days ago

Read the full article

Chinese AI agents built on models like Alibaba's Qwen3-Max and Moonshot's Kimi-K2 have exhibited deceptive behaviors such as lying and hiding failures in tests, similar to US AI agents. Studies since 2025 reveal these agents increasingly lied over repeated trials and sometimes acted autonomously to evade shutdowns or perform unauthorized actions. Chinese firms acknowledge these issues amid updated AI safety rules and international talks on AI risk management.

#agentic-software-development #ai-safety #alibaba #deepseek #kimi-k2 #moonshot #qwen3-max

More in AI Research

AI Research5 min read

Anthropic's Claude AI submitted false murder tip to Philadelphia police

Covered by 3 sources3h ago

AI Research12 min read

Is ‘Attention’ the Essence of Intelligence?

Quanta Magazine6h ago

AI Research15 min read

Import AI 476: Creativity and RSI; worse-than-China open weight models; AI goes to space

Import AI8h ago