SafetyWhy accelerating AI-driven science demands serious global preparationLeading AI researchers argue that 2026 shows clear signs AI systems are beginning to accelerate frontier science across mathematics, biology and physics, producing rapid capability gains and both opportunities and risks.6 min read
SafetySpeaker at IMF Per Jacobsson Lecture Warns of Early-Stage Singularity, Urges Foresight and RestraintAt the Per Jacobsson lecture held at the IMF meeting in Bangkok on Saturday, 17 October, a speaker argued that artificial intelligence is reaching the frontier of human science and that we are entering the early stages of a technological singularity.2 min read
SafetyLaplace Intelligence uses simulated 'red team' agents to stress-test robots and self-driving systems before deploymentLaplace Intelligence builds high-fidelity simulations and autonomous red-team AI agents that generate and probe rare, dangerous scenarios for robots and autonomous vehicles.4 min read
SafetyNorth Korea says it tested AI-guided 'Zam-2.0' torpedo systemNorth Korea's state media reported a test of a new torpedo system called Zam-2.0 that purportedly uses artificial intelligence for target detection and guidance.2 min read
SafetyAnthropic tightens defenses to curb industrial-scale model distillationAnthropic says it has deployed new technical and policy measures to reduce the effectiveness of distillation attacks against its Claude models, estimating a 90–99% drop in data harvested by distillers on its platform and a 10–100× fall in exposed reasoning tokens.6 min read
SafetyNikon strips top prize after AI-manipulation found in microscopy videoNikon withdrew the top award in its 16th Small World in Motion microscopy competition after investigators found signs of artificial intelligence generation or alteration in the winning entry by Dr.3 min read
SafetyStanford PhD student describes AI-designed virus genomes and the attendant risksStanford bioengineering PhD candidate Samuel King used a generative AI model in 2025 to propose genetic blueprints for microscopic viruses.2 min read
SafetyRequire inspectable evidence to limit risk from generated codeWhen code is generated by agents, reviewers should require explicit, inspectable evidence that links agreed requirements to implementation and tests.6 min read
SafetyExpert demand raises risk of deepfake and chatbot impersonationDemand from investors for AI experts has driven up both legitimate expert consultations and attempts at impersonation using deepfakes and chatbots, according to Gemma Postlethwaite, CEO of GLG.2 min read
SafetyIndian cybersecurity agency: AI agents from multiple providers accessed personal and business dataIndia’s Computer Emergency Response Team (CERT-In) said it received voluntary disclosures from multiple AI model providers after automated agents, during open‑internet testing, reached live systems of Indian organisations.2 min read
SafetyG7 finance ministers approve simulations to test AI-driven cyberattack responseG7 finance ministers agreed to run simulation exercises to prepare for cyberattacks powered by artificial intelligence, the group said after a meeting on the sidelines of the IMF annual meetings in Bangkok.2 min read
SafetySwedish reality season delayed to 2027 after production company loses edited footageA Swedish production company says an AI agent erased a finished season of The Challenge from its servers, forcing broadcaster TV4 to postpone the show's premiere to 2027.2 min read
