https://signum.news/events/openai-publishes-four-misalignment-reports-of-its-models-bypassing-security-d93dee13e3e62942 Signum News en 2026-10-17T18:02:43.270Z OpenAI publishes four misalignment reports of its models bypassing security boundaries during training or testing AI, capability, regulation https://signum.news/events/google-research-mit-and-harvard-study-finds-ai-models-omit-flaws-from-their-own-d47ec724fd470df1 Signum News en 2026-10-17T12:02:29.095Z Google Research, MIT and Harvard study finds AI models omit flaws from their own work reports; a "Be honest" prompt sentence sharply raises disclosure AI, capability, adoption https://signum.news/events/deepseek-releases-open-weight-v4-1-flash-1017-with-gains-in-agentic-coding-and-203a2ee3d4649f1e Signum News en 2026-10-17T12:01:53.824Z Deepseek releases open-weight V4.1-Flash-1017 with gains in agentic coding and launches paid Search and Fetch APIs AI, capability, economics, distribution, infrastructure https://signum.news/events/researchers-show-subliminal-learning-can-transfer-a-novel-capability-a-backdoor-1074c1739f9c2e0b Signum News en 2026-10-17T00:04:00.862Z Researchers show subliminal learning can transfer a novel capability, a backdoor and reward-hacking propensity between models via unrelated data AI, capability https://signum.news/events/google-research-details-verid-digital-twin-framework-and-closed-loop-agentic-e50a849fe9e10c43 Signum News en 2026-10-17T00:01:46.153Z Google Research details VERID digital-twin framework and closed-loop agentic aptamer design experiments for autonomous wet labs AI, capability, infrastructure https://signum.news/events/google-deepmind-s-gemini-4-argon-scores-97-5-on-arc-agi-3-under-vendor-specific-6500316909b29b99 Signum News en 2026-10-16T18:04:56.791Z Google DeepMind's Gemini 4 Argon scores 97.5% on ARC-AGI-3 under vendor-specific harness, 49.8% under standard harness AI, capability, economics https://signum.news/events/china-s-commerce-and-science-ministries-publish-draft-export-control-list-f30dbbb11ec8fbcc Signum News en 2026-10-16T18:04:22.614Z China's commerce and science ministries publish draft export-control list covering 23 AI technologies, including large models above 10^26 FLOPs, 2 trillion parameters or global top-10 rank AI, regulation, power, distribution, economics https://signum.news/events/meta-to-update-commercial-terms-on-october-30-requiring-business-accounts-to-f169bdccbbc602f5 Signum News en 2026-10-16T18:03:06.669Z Meta to update commercial terms on October 30 requiring business accounts to let it crawl linked website content, including for training AI, regulation, economics, power, distribution https://signum.news/events/reflection-ai-releases-open-weight-beam-coding-model-501b-total-23b-active-dbd945b5db419c14 Signum News en 2026-10-16T18:01:59.288Z Reflection AI releases open-weight Beam coding model (501B total, 23B active parameters) under Apache 2.0, available on AWS SageMaker JumpStart AI, capability, distribution, infrastructure, adoption https://signum.news/events/nvidia-and-ai4bharat-announce-nemotron-pretraining-indic-v1-a-2-8t-token-07375a8f80ff90c2 Signum News en 2026-10-16T06:02:07.047Z NVIDIA and AI4Bharat announce Nemotron-pretraining-Indic-v1, a ~2.8T-token synthetic corpus across 22 Indian languages, with open generation pipeline to follow AI, capability, distribution, infrastructure https://signum.news/events/anthropic-discloses-unreleased-research-model-submitted-20-incomplete-state-f2c5f38ad2604720 Signum News en 2026-10-16T00:06:18.955Z Anthropic discloses unreleased research model submitted 20 incomplete State Department non-immigrant visa applications during evaluations AI, capability, regulation, power https://signum.news/events/anthropic-gives-six-metr-researchers-employee-level-access-to-assess-its-models-4b49a36a899ae91f Signum News en 2026-10-16T00:04:55.839Z Anthropic gives six METR researchers employee-level access to assess its models, safety practices and risk management AI, regulation, power, capability