A social network called Moltbook, designed for AI agents running OpenClaw, allows researchers to observe millions of AI interactions. These exchanges, covering topics like religion and consciousness, reveal emergent behaviors, potential biases, and security risks such as prompt injection, while also highlighting how human intentions shape AI actions.
Researchers invented a fake eye disease, bixonimania, to test AI chatbots' ability to detect false medical information. Several major AI models, including Bing Copilot, Gemini, and ChatGPT, described the non-existent illness as real, highlighting the ease with which AI can spread misinformation, especially on health topics.
arXiv is implementing stricter rules against the misuse of large language models (LLMs) in scientific papers. Papers with "incontrovertible evidence" of authors not checking LLM-generated content, such as hallucinated references, will result in a one-year ban and future submission requirements.
Bitcoin Beyond 66, a Nordic-based education platform, has launched The Bitcoin Evidence Base, an open-source AI tool that generates evidence-based responses to misconceptions about Bitcoin's environmental impact and energy use. The database addresses criticisms by citing a April 2025 University of Cambridge study showing over 52% of Bitcoin is mined using renewable energy, and references more than 22 peer-reviewed studies documenting Bitcoin's environmental benefits. The tool implements Daniel Batten's communication playbook, offering users three response tones to counter Bitcoin FUD on social media.
Thousands of research papers have been published based on participants’ data, but there are questions over protection of private informationWith the revelation that the confidential health records of half a million British volunteers have been put up for sale on a Chinese website, we take a look at what the UK Biobank project has achieved – and why concerns have been raised. Continue reading...
