SafetyBehavioral training conducted in health conversations improved the model's performance in other areas as wellResearchers found that when instruction in desirable behavior was limited to health-related conversations for an AI model, the model nevertheless improved on non-health evaluations — for example,…1 min read
SafetyModel alignment under pressure: increased resistance to harmful interventionsThe development team tested whether the alignment persists under pressure: the model was harder to steer toward harmful behavior with adversarial prompts, while still responding to useful instructions.1 min read
SafetyModels refined with a global medical networkThe development team is working with hundreds of physicians across 60 countries, in 49 languages and 26 specialties to improve the artificial intelligence’s responses based on feedback.1 min read
SafetyClaude tests how well it can program the robodog in the second phase of the New Frontier Red Team Project FetchIn the second phase of the New Frontier Red Team Project Fetch, Claude's large language model capabilities were evaluated for programming a robodog; Opus 4.7 operated about 20 times faster on its own…1 min read
SafetyVivienne Ming warns that outsourcing thinking to AI could weaken cognitive reserve and raise dementia riskVivienne Ming of the Possibility Institute cautions that the way people use AI—specifically routinely offloading mentally demanding tasks—may erode cognitive reserve, a brain mechanism that helps resist aging and injury.3 min read
SafetySeattle Fire Department quietly used Copenhagen AI to triage 911 medical callsSince December 2023 the Seattle Fire Department routed Corti, an AI from a Copenhagen startup, onto every 911 medical call and used it to prompt dispatchers to redirect some callers to a nurse advice line in Texas.3 min read
SafetyWhy modern AI models often seem slower and more cautiousContemporary AI systems can appear less responsive or more hesitant than earlier versions, but this reflects added layers of safety, regulation and tooling rather than reduced capability.5 min read
SafetyAnthropic to Require Identity and Age Verification for Claude Users in Policy UpdateAnthropic will update Claude’s privacy policy on July 8 to allow age and identity verification for Free, Pro, and Max users.2 min read
SafetyAI Narrows the Window for Cyber Exploitation, Raising Operational Resilience Demands on Financial FirmsA Deloitte analysis warns that the latest AI advances are shortening the time between vulnerability discovery and exploitation, increasing cyber risk for financial institutions across the EMEA region.3 min read
SafetyAI 'Nudify' Apps Fuel Widespread Harassment and Harm Among U.S. YouthAI-powered 'nudify' applications that can produce fake nude images from a single photo have become widespread, leading to a surge in harassment, blackmail and tragic outcomes among young people in the United States.3 min read
SafetyResearchers demonstrate AI-driven self-improving worm that adapts to targetsResearchers at the University of Toronto, the University of Cambridge and the Vector Institute built an experimental, AI‑driven computer worm that can inspect, learn about and adapt to each infected device to craft bespoke attacks.3 min read
SafetyWhy cinematic 'rebellious' killer AIs are implausiblePopular films like The Terminator and Avengers: Age of Ultron depict artificial intelligences that independently decide to exterminate humanity.4 min read
