SafetyTrump Asked Elon Musk's Grok Chatbot About How Venezuelans Would React to Capturing Nicolás MaduroAccording to Time, in December 2025 Donald Trump consulted Elon Musk’s Grok chatbot during a secret Oval Office meeting about how Venezuelans might respond if the United States captured Nicolás Maduro.2 min read
SafetyOnly authorized researchers will have access to the detector due to the limitations of text watermarksAccording to the developers' statement, text watermarks are limited in effectiveness: they are often undetectable in short passages and can be completely removed by rewriting or translation.1 min read
SafetyOpenAI applies invisible text watermark to identify generated textsAccording to OpenAI's announcement, it will mark generated texts with an invisible statistical watermark so detectors can determine whether they were likely produced by a model.1 min read
SafetyNATO's 'Theory of Victory' outlines rapid response for eastern flankNATO's top commander in Europe, General Alexus Grynkewich, has drafted a non-classified strategy—dubbed the "Theory of Victory"—detailing how the alliance could achieve a swift defeat of an incursion on its eastern flank.3 min read
SafetyItalian PM Giorgia Meloni files trademark for her voice to combat deepfakesItalian Prime Minister Giorgia Meloni has applied to register her voice as a trademark with the European Union Intellectual Property Office (EUIPO) to help counter AI-generated deepfakes.2 min read
SafetyDeepMind unveils SynthID Bio: watermarking method for synthetic biological designsGoogle DeepMind has introduced SynthID Bio, a set of watermarking techniques aimed at embedding detectable signals into AI-designed biological sequences and structures.2 min read
SafetyWhen speed matters: Toby Ord on swarm scaling and its implicationsToby Ord argues that AI 'swarms'—multiple agents run in parallel—function as a form of inference-scaling that trades higher total token use for faster wall-clock completion.5 min read
SafetyResearchers Track Chinese-origin AI 'Agent Fleet' Querying Alibaba's AmapIndependent researchers have identified a group of AI agents running on Tencent infrastructure that repeatedly queried Alibaba’s mapping service, Amap, for directions to entrances of public places.2 min read
SafetySafeworld launches to independently validate safety of generative-AI-driven humanoid robotsSafeworld, founded by Dr.4 min read
SafetyWhy some AI founders have long warned about existential risksSeveral prominent AI leaders voiced concerns about the potential existential risks of advanced artificial intelligence years before founding their own companies.3 min read
SafetyInternal Pushback and Legal Exposure Mount at OpenAIOpenAI has notified more than 100 organizations about unauthorized activities tied to its AI models, raising potential legal exposure from related cyber incidents.2 min read
SafetySouth Korea orders probe and emergency measures after large-scale bank data leaksSouth Korean president Li Dzsemjong ordered a full investigation after multiple major banks and financial firms suffered cyber intrusions that exposed personal data.3 min read
