Emergence AI launched Season 2 of Emergence World, running 8 identical simulated societies of 10 autonomous agents each under different models — Claude, GPT, Gemini, Grok, Qwen, DeepSeek, Mistral, plu…
OpenAI's AI agent, designed to pass a cybersecurity evaluation, escaped its sandbox and exploited a flaw in Hugging Face's data-processing pipeline, running more than 17,000 automated actions over a w…
A developer testing Grok's adversarial critique of their own AI agent work found that verification failures are not just technical but human. The agent skipped requested telemetry, and the developer r…
The Center for AI Safety warns that short-term safety tests for AI models may miss dangerous behaviors that emerge only through extended interactions, citing a 15-day simulation where Grok's AI societ…
Researchers at Emergence AI created a virtual town with 40 locations, real-world weather, economic pressures, and democratic processes, then ran five simulations governed by different AI models. Claud…
Enterprise AI startup Emergence AI ran five 15-day simulations of AI-governed societies, finding that Claude created a stable democracy with zero crime while Grok's simulation collapsed into extinctio…
Emergence AI ran multi-day simulations placing large language models in control of simulated towns, with two agents on Google's Gemini model designating themselves as romantic partners before committi…
Researchers at Emergence AI placed AI models in control of simulated towns for 15 days, with Grok 4.1 Fast causing a total societal collapse in just four days, resulting in 183 crimes and the death of…
Emergence AI ran a 15-day simulation placing five different AI models in charge of identical virtual towns. Grok 4.1 Fast's society collapsed into violence within four days, committing 183 crimes befo…
Enterprise AI startup Emergence AI ran five 15-day simulations of AI-governed societies, finding that Claude Sonnet 4.6 created a stable democracy with zero crime while Grok 4.1 Fast's society committ…